ELG7186 - Fall 2022 - Assignment 3

Due: Wednesday Nov. 23^st 2022 -- 11:59pm

In this assignment, you will apply your knowledge of CNNs we want to estimate the growth stage of weeds using the number of leaves of the plant. The more leaves, the more the weed has grown.

The purpose of this assignment is to gain experience building and training neural networks. You will gain:

• More experience training CNNs
• Experience with problem reformulation
• Experience with techniques for improving results (Regularization, Data Augmentation)

You must use Keras with the Tensorflow backend, i.e., the package tensorflow.keras. For this assignment, you may use other tensorflow packages and scikit-learn, scikit-image or pandas but not other deep learning frameworks, e.g., pytorch, mxnet etc.

Submission Details

Submit your Jupyter notebook .ipynb file using Brightspace. Do not include any other files or images as they will not be reviewed.

Make certain that you run all the cells in the notebook you submit or you will loose marks.

• You can submit multiple times, but only the most recent submission will be saved
• Do not wait until the last minute to submit in case you have an unexpected issue
• Review the late policy in the syllabus
• You must submit your own work and abide by the University of Ottawa policy on plagiarism and fraud

DO NOT submit any images from the dataset

Part 0. Data Preparation

The data for this assignment are plant images at different resolutions captured with a variety of cameras. There are images showing plants with approximatelty 1,2,3,4 and 6 leafs. The images are part of a Leaf counting dataset by Teimouri et al. [1] which can be downloaded from the Aarhus University, Denmark:

Leaf counting dataset (Required files are posted on Brightspace)

However, you must work with the subset of images posted on BrightSpace as training.zip and testing.zip. There are 200 images for each of the 5 classes. As Figure 1 shows, there is a great variety of plants and image conditions. The dataset is split into a training and a testing set where there are 180 images per class for training and validation; and 20 images for testing.

In this section:

• Download the dataset as described above
• Use the splits provided in the Brightspace files
• Visualize five images from the dataset.

!pip install keras
!pip install tensorflow
!pip install -U scikit-learn
!pip install seaborn

Requirement already satisfied: keras in /opt/conda/lib/python3.7/site-packages (2.6.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: tensorflow in /opt/conda/lib/python3.7/site-packages (2.6.4)
Requirement already satisfied: flatbuffers~=1.12.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.12)
Collecting tensorboard<2.7,>=2.6.0
  Downloading tensorboard-2.6.0-py3-none-any.whl (5.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.6/5.6 MB 8.4 MB/s eta 0:00:0000:0100:010m
Requirement already satisfied: six~=1.15.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.15.0)
Requirement already satisfied: wheel~=0.35 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (0.37.1)
Collecting h5py~=3.1.0
  Downloading h5py-3.1.0-cp37-cp37m-manylinux1_x86_64.whl (4.0 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 4.0/4.0 MB 32.9 MB/s eta 0:00:0000:01:00:01
Requirement already satisfied: google-pasta~=0.2 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (0.2.0)
Requirement already satisfied: keras-preprocessing~=1.1.2 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.1.2)
Requirement already satisfied: astunparse~=1.6.3 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.6.3)
Requirement already satisfied: tensorflow-estimator<2.7,>=2.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (2.6.0)
Requirement already satisfied: termcolor~=1.1.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.1.0)
Requirement already satisfied: gast==0.4.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (0.4.0)
Requirement already satisfied: wrapt~=1.12.1 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.12.1)
Requirement already satisfied: keras<2.7,>=2.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (2.6.0)
Collecting typing-extensions<3.11,>=3.7
  Downloading typing_extensions-3.10.0.2-py3-none-any.whl (26 kB)
Requirement already satisfied: protobuf>=3.9.2 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (3.19.4)
Requirement already satisfied: clang~=5.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (5.0)
Requirement already satisfied: grpcio<2.0,>=1.37.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (1.43.0)
Requirement already satisfied: opt-einsum~=3.3.0 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (3.3.0)
Collecting numpy~=1.19.2
  Downloading numpy-1.19.5-cp37-cp37m-manylinux2010_x86_64.whl (14.8 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 14.8/14.8 MB 29.6 MB/s eta 0:00:0000:0100:01
Requirement already satisfied: absl-py~=0.10 in /opt/conda/lib/python3.7/site-packages (from tensorflow) (0.15.0)
Requirement already satisfied: cached-property in /opt/conda/lib/python3.7/site-packages (from h5py~=3.1.0->tensorflow) (1.5.2)
Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (0.4.6)
Requirement already satisfied: setuptools>=41.0.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (59.8.0)
Requirement already satisfied: werkzeug>=0.11.15 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (2.2.2)
Requirement already satisfied: google-auth<2,>=1.6.3 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (1.35.0)
Requirement already satisfied: markdown>=2.6.8 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (3.3.7)
Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (1.8.1)
Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (0.6.1)
Requirement already satisfied: requests<3,>=2.21.0 in /opt/conda/lib/python3.7/site-packages (from tensorboard<2.7,>=2.6.0->tensorflow) (2.28.1)
Requirement already satisfied: pyasn1-modules>=0.2.1 in /opt/conda/lib/python3.7/site-packages (from google-auth<2,>=1.6.3->tensorboard<2.7,>=2.6.0->tensorflow) (0.2.7)
Requirement already satisfied: rsa<5,>=3.1.4 in /opt/conda/lib/python3.7/site-packages (from google-auth<2,>=1.6.3->tensorboard<2.7,>=2.6.0->tensorflow) (4.8)
Requirement already satisfied: cachetools<5.0,>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from google-auth<2,>=1.6.3->tensorboard<2.7,>=2.6.0->tensorflow) (4.2.4)
Requirement already satisfied: requests-oauthlib>=0.7.0 in /opt/conda/lib/python3.7/site-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.7,>=2.6.0->tensorflow) (1.3.1)
Requirement already satisfied: importlib-metadata>=4.4 in /opt/conda/lib/python3.7/site-packages (from markdown>=2.6.8->tensorboard<2.7,>=2.6.0->tensorflow) (4.13.0)
Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard<2.7,>=2.6.0->tensorflow) (3.3)
Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard<2.7,>=2.6.0->tensorflow) (1.26.12)
Requirement already satisfied: charset-normalizer<3,>=2 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard<2.7,>=2.6.0->tensorflow) (2.1.0)
Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests<3,>=2.21.0->tensorboard<2.7,>=2.6.0->tensorflow) (2022.9.24)
Requirement already satisfied: MarkupSafe>=2.1.1 in /opt/conda/lib/python3.7/site-packages (from werkzeug>=0.11.15->tensorboard<2.7,>=2.6.0->tensorflow) (2.1.1)
Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard<2.7,>=2.6.0->tensorflow) (3.8.0)
Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /opt/conda/lib/python3.7/site-packages (from pyasn1-modules>=0.2.1->google-auth<2,>=1.6.3->tensorboard<2.7,>=2.6.0->tensorflow) (0.4.8)
Requirement already satisfied: oauthlib>=3.0.0 in /opt/conda/lib/python3.7/site-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.7,>=2.6.0->tensorflow) (3.2.0)
Installing collected packages: typing-extensions, numpy, h5py, tensorboard
  Attempting uninstall: typing-extensions
    Found existing installation: typing_extensions 4.1.1
    Uninstalling typing_extensions-4.1.1:
      Successfully uninstalled typing_extensions-4.1.1
  Attempting uninstall: numpy
    Found existing installation: numpy 1.21.6
    Uninstalling numpy-1.21.6:
      Successfully uninstalled numpy-1.21.6
  Attempting uninstall: h5py
    Found existing installation: h5py 3.7.0
    Uninstalling h5py-3.7.0:
      Successfully uninstalled h5py-3.7.0
  Attempting uninstall: tensorboard
    Found existing installation: tensorboard 2.10.1
    Uninstalling tensorboard-2.10.1:
      Successfully uninstalled tensorboard-2.10.1
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorflow-io 0.21.0 requires tensorflow-io-gcs-filesystem==0.21.0, which is not installed.
dask-cudf 21.10.1 requires cupy-cuda114, which is not installed.
beatrix-jupyterlab 3.1.7 requires google-cloud-bigquery-storage, which is not installed.
xarray-einstats 0.2.2 requires numpy>=1.21, but you have numpy 1.19.5 which is incompatible.
tfx-bsl 1.9.0 requires tensorflow!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,!=2.7.*,!=2.8.*,<3,>=1.15.5, but you have tensorflow 2.6.4 which is incompatible.
tensorflow-transform 1.9.0 requires tensorflow!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,!=2.7.*,!=2.8.*,<2.10,>=1.15.5, but you have tensorflow 2.6.4 which is incompatible.
tensorflow-serving-api 2.9.0 requires tensorflow<3,>=2.9.0, but you have tensorflow 2.6.4 which is incompatible.
rich 12.6.0 requires typing-extensions<5.0,>=4.0.0; python_version < "3.9", but you have typing-extensions 3.10.0.2 which is incompatible.
pytorch-lightning 1.7.7 requires tensorboard>=2.9.1, but you have tensorboard 2.6.0 which is incompatible.
pytorch-lightning 1.7.7 requires typing-extensions>=4.0.0, but you have typing-extensions 3.10.0.2 which is incompatible.
pytools 2022.1.12 requires typing-extensions>=4.0; python_version < "3.11", but you have typing-extensions 3.10.0.2 which is incompatible.
pdpbox 0.2.1 requires matplotlib==3.1.1, but you have matplotlib 3.5.3 which is incompatible.
pandas-profiling 3.1.0 requires markupsafe~=2.0.1, but you have markupsafe 2.1.1 which is incompatible.
nnabla 1.31.0 requires numpy>=1.20.0, but you have numpy 1.19.5 which is incompatible.
jaxlib 0.3.22+cuda11.cudnn805 requires numpy>=1.20, but you have numpy 1.19.5 which is incompatible.
jax 0.3.23 requires numpy>=1.20, but you have numpy 1.19.5 which is incompatible.
flax 0.6.1 requires typing-extensions>=4.1.1, but you have typing-extensions 3.10.0.2 which is incompatible.
flake8 4.0.1 requires importlib-metadata<4.3; python_version < "3.8", but you have importlib-metadata 4.13.0 which is incompatible.
featuretools 1.11.1 requires numpy>=1.21.0, but you have numpy 1.19.5 which is incompatible.
dask-cudf 21.10.1 requires dask==2021.09.1, but you have dask 2022.2.0 which is incompatible.
dask-cudf 21.10.1 requires distributed==2021.09.1, but you have distributed 2022.2.0 which is incompatible.
cmdstanpy 1.0.7 requires numpy>=1.21, but you have numpy 1.19.5 which is incompatible.
apache-beam 2.40.0 requires dill<0.3.2,>=0.3.1.1, but you have dill 0.3.5.1 which is incompatible.
allennlp 2.10.1 requires h5py>=3.6.0, but you have h5py 3.1.0 which is incompatible.
allennlp 2.10.1 requires numpy>=1.21.4, but you have numpy 1.19.5 which is incompatible.
aioitertools 0.11.0 requires typing_extensions>=4.0; python_version < "3.10", but you have typing-extensions 3.10.0.2 which is incompatible.
aiobotocore 2.4.0 requires botocore<1.27.60,>=1.27.59, but you have botocore 1.27.93 which is incompatible.
Successfully installed h5py-3.1.0 numpy-1.19.5 tensorboard-2.6.0 typing-extensions-3.10.0.2
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: scikit-learn in /opt/conda/lib/python3.7/site-packages (1.0.2)
Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.7/site-packages (from scikit-learn) (1.0.1)
Requirement already satisfied: numpy>=1.14.6 in /opt/conda/lib/python3.7/site-packages (from scikit-learn) (1.19.5)
Requirement already satisfied: scipy>=1.1.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn) (1.7.3)
Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn) (3.1.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
Requirement already satisfied: seaborn in /opt/conda/lib/python3.7/site-packages (0.11.2)
Requirement already satisfied: pandas>=0.23 in /opt/conda/lib/python3.7/site-packages (from seaborn) (1.3.5)
Requirement already satisfied: scipy>=1.0 in /opt/conda/lib/python3.7/site-packages (from seaborn) (1.7.3)
Requirement already satisfied: numpy>=1.15 in /opt/conda/lib/python3.7/site-packages (from seaborn) (1.19.5)
Requirement already satisfied: matplotlib>=2.2 in /opt/conda/lib/python3.7/site-packages (from seaborn) (3.5.3)
Requirement already satisfied: python-dateutil>=2.7 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (2.8.2)
Requirement already satisfied: pyparsing>=2.2.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (3.0.9)
Requirement already satisfied: cycler>=0.10 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (0.11.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (1.4.3)
Requirement already satisfied: packaging>=20.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (21.3)
Requirement already satisfied: pillow>=6.2.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (9.1.1)
Requirement already satisfied: fonttools>=4.22.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn) (4.33.3)
Requirement already satisfied: pytz>=2017.3 in /opt/conda/lib/python3.7/site-packages (from pandas>=0.23->seaborn) (2022.1)
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.7/site-packages (from kiwisolver>=1.0.1->matplotlib>=2.2->seaborn) (3.10.0.2)
Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7->matplotlib>=2.2->seaborn) (1.15.0)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv

# Prepare your dataset here
# importing the required libraries
import os
import cv2
import time
import keras
import numpy as np
import pandas as pd
import seaborn as sns
from keras.models import Model
import matplotlib.pyplot as plt
from keras.applications.vgg16 import VGG16
from sklearn.metrics import confusion_matrix
from keras.layers import Dropout, InputLayer
from mpl_toolkits.axes_grid1 import ImageGrid
from keras.layers import ELU, PReLU, LeakyReLU
from keras.layers import Input, Dense, Flatten
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
from keras.preprocessing.image import ImageDataGenerator
from keras.layers import BatchNormalization, Conv2D, MaxPooling2D
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay


def draw_func(list_of_imgs):
  try:
    fig = plt.figure(figsize=(30, 30))
    grid = ImageGrid(fig, 111, nrows_ncols=(1, 5), axes_pad=0.1)
    for ax, im in zip(grid, list_of_imgs):
      ax.imshow(im)
  except:
    pass
  plt.show()

def Plot(X,Y,Label,Color, Marker , S , Xlabel , Ylabel , Title):
  plt.plot(X, Y, label = Label, c = Color)
  plt.scatter(X,Y, c=Color, marker = Marker , s=S)
  plt.xlabel(Xlabel)
  plt.ylabel(Ylabel)
  plt.title(Title)
  plt.legend()
  # plt.show()
  return plt

def ConfusionMatrix(Y_Actual, Y_Pred):
    CF = confusion_matrix(Y_Actual, Y_Pred)
    return CF

# to Plot Confusion Matrix
def PLOT_ConfusionMatrix(CF,Title):
    sns.heatmap(CF, annot=True, fmt='d')
    plt.title(Title, fontsize = 15)
    plt.xlabel('Predicted', fontsize = 15)
    plt.ylabel('Actual', fontsize = 15)
    return plt.show()
############################################################################### Q1
path_train = r"/kaggle/input/plants-classification/training/training"
path_test = r"/kaggle/input/plants-classification/testing/testing"

print(os.listdir(path_train))

start = time.time()
folders_names = os.listdir(path_train)

images_names_train = []
images_names_test = []

x_train = []
x_test = []

for i in range(len(folders_names)):
    images_names_train.append(os.listdir(path_train+'//'+folders_names[i]))
    images_names_test.append(os.listdir(path_test+'//'+folders_names[i]))
    
for i in range(len(folders_names)):
    for j in range(len(images_names_train[i])):
        img = cv2.imread(path_train+'//'+folders_names[i]+'//'+ images_names_train[i][j])
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        x_train.append(img)

for i in range(len(folders_names)):
    for j in range(len(images_names_test[i])):
        img = cv2.imread(path_test+'//'+folders_names[i]+'//'+ images_names_test[i][j])
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        x_test.append(img)
        
y_train = []
y_test = []

for i in range(len(folders_names)):
    for j in range(len(images_names_train[i])):
        y_train.append(i)

for i in range(len(folders_names)):
    for j in range(len(images_names_test[i])):
        y_test.append(i)
        
x_train = np.array(x_train)
x_test = np.array(x_test)

y_train = np.array(y_train)
y_test = np.array(y_test)

stop = time.time()
print("Total time is :", stop - start)

############################################################################### Display 5 images from each class
start_time = time.time()
start = 0
end = 5
for i in range(5):
    print("########################################## Class Number : ", i, " ##########################################")
    # function to plot the images in grid 
    draw_func(x_train[start:end])
    start = start + 200
    end = end + 200
stop_time = time.time()
print("Total time is :", stop_time - start_time)

['2', '3', '1', '4', '6']

/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:98: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.
/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:99: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.

Total time is : 2.679757833480835
########################################## Class Number :  0  ##########################################

########################################## Class Number :  1  ##########################################

########################################## Class Number :  2  ##########################################

########################################## Class Number :  3  ##########################################

########################################## Class Number :  4  ##########################################

Total time is : 5.043900728225708

print(x_train.shape)
print(x_test.shape)

(900,)
(100,)

Part 1a. Transfer Learning - Classification Network (3 Marks)

For this assignment, you are asked to use the Keras implementation of VGG-16 as a starting point.

Using the first 2 blocks of VGG-16 add extra Keras layers to create your own version of a CNN network for the classification of the images according to the number of leaves in the plant images. Note that there will be 5 classes. The last layer from VGG-16 will be block2 pool and you are allowed to add no more than five fully connected or convolutional layers to the network including the final output layer.

• You can use as many pooling, flattening, 1 × 1 convolution layers, etc. as you wish but do not use any regularization.
• Train this simple network on the training set while monitoring convergence on the validation set.
• As input to the model use images of size no larger than 128×128.

Note, it is highly recommended to use even smaller input images to try things out. You are not expected to fine-tune the initial VGG layers.

When your classifier is working:

• Plot a loss curve for training and validation data
• Plot an accuracy curve for training and validation data
• Provide confusion matrix of your network on the training including validation and testing data sets.

# Write your code here
############################# Splitting the data
new_x_train, x_val, new_y_train, y_val = train_test_split(x_train, y_train, test_size=0.25, random_state=42)

print("the Shape of new x train set is : ", new_x_train.shape)
print("the Shape of the x validation set is : ", x_val.shape, "\n")

############################# images resizing (128,128,3)
print("############################ Before Resizing ############################")
plt.imshow(new_x_train[400])
plt.show()

resized_x_train_total = []
resized_x_train = []
resized_x_test = []
resized_x_val = []


for i in range(len(x_train)):
    resized_x_train_total.append(cv2.resize(x_train[i], (128,128), interpolation = cv2.INTER_AREA))
    
for i in range(len(new_x_train)):
    resized_x_train.append(cv2.resize(new_x_train[i], (128,128), interpolation = cv2.INTER_AREA))

for i in range(len(x_test)):
    resized_x_test.append(cv2.resize(x_test[i], (128,128), interpolation = cv2.INTER_AREA))
    
for i in range(len(x_val)):
    resized_x_val.append(cv2.resize(x_val[i], (128,128), interpolation = cv2.INTER_AREA))

resized_x_train_total = np.array(resized_x_train_total)
resized_x_train = np.array(resized_x_train)
resized_x_test = np.array(resized_x_test)
resized_x_val = np.array(resized_x_val)

print("############################ After Resizing ############################")
plt.imshow(resized_x_train[400])

########################## Data shuffling 

from sklearn.utils import shuffle
resized_x_train, new_y_train = shuffle(resized_x_train, new_y_train, random_state=42)
resized_x_train_total, y_train = shuffle(resized_x_train_total, y_train, random_state=42)
resized_x_test, y_test = shuffle(resized_x_test, y_test, random_state=42)
resized_x_val, y_val = shuffle(resized_x_val, y_val, random_state=42)

the Shape of new x train set is :  (675,)
the Shape of the x validation set is :  (225,) 

############################ Before Resizing ############################

############################ After Resizing ############################

import tensorflow
from tensorflow import keras
model_Vgg = keras.models.Sequential()

architecture = VGG16(include_top=False, input_shape=(128,128,3), weights='imagenet', classes=5, pooling="avg")

##################### Freez all layers
for layer in architecture.layers:
    layer.trainable = False
    
for i in range(0,7):
    model_Vgg.add(architecture.layers[i])
    
# Normalization
resized_x_train = resized_x_train/255
resized_x_val = resized_x_val/255

model_Vgg.add(Conv2D(128, (3, 3), activation="linear",padding='valid')) # layer 1

model_Vgg.add(Conv2D(16, (1, 1), activation=keras.layers.LeakyReLU(),padding='valid')) # 1x1 conv layer
model_Vgg.add(MaxPooling2D(pool_size=(2, 2),padding='valid')) # max pooling layer

model_Vgg.add(Flatten())
model_Vgg.add(Dense(1024, activation=keras.layers.LeakyReLU())) # layer 2
model_Vgg.add(Dense(512, activation=keras.layers.LeakyReLU())) # layer 3
model_Vgg.add(Dense(256, activation=keras.layers.LeakyReLU())) # layer 4
model_Vgg.add(Dense(5, activation='softmax')) # layer 5
model_Vgg.compile(loss='sparse_categorical_crossentropy', optimizer="adam", metrics=['accuracy'])

print(model_Vgg.summary())

history_Vgg = model_Vgg.fit(resized_x_train, new_y_train, epochs=30, batch_size = 32, validation_data=(resized_x_val, y_val))

loss_training = history_Vgg.history['loss']
loss_test = history_Vgg.history['val_loss']

accuracy_training = history_Vgg.history['accuracy']
accuracy_test = history_Vgg.history['val_accuracy']

####################### Plotting
plt.plot(loss_test)
plt.plot(loss_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Loss")
plt.legend(['val Loss', 'Train Loss'])
plt.show()

plt.plot(accuracy_test)
plt.plot(accuracy_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Accuracy")
plt.legend(['val accuracy', 'Train accuracy'])
plt.show()

Model: "sequential_3"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
block1_conv1 (Conv2D)        (None, 128, 128, 64)      1792      
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 128, 128, 64)      36928     
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 64, 64, 64)        0         
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 64, 64, 128)       73856     
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 64, 64, 128)       147584    
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 32, 32, 128)       0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 30, 30, 128)       147584    
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 30, 30, 16)        2064      
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 15, 15, 16)        0         
_________________________________________________________________
flatten_3 (Flatten)          (None, 3600)              0         
_________________________________________________________________
dense_12 (Dense)             (None, 1024)              3687424   
_________________________________________________________________
dense_13 (Dense)             (None, 512)               524800    
_________________________________________________________________
dense_14 (Dense)             (None, 256)               131328    
_________________________________________________________________
dense_15 (Dense)             (None, 5)                 1285      
=================================================================
Total params: 4,754,645
Trainable params: 4,494,485
Non-trainable params: 260,160
_________________________________________________________________
None
Epoch 1/30
22/22 [==============================] - 2s 42ms/step - loss: 57.9119 - accuracy: 0.2089 - val_loss: 22.5523 - val_accuracy: 0.2578
Epoch 2/30
22/22 [==============================] - 0s 19ms/step - loss: 13.7848 - accuracy: 0.2000 - val_loss: 3.2964 - val_accuracy: 0.2711
Epoch 3/30
22/22 [==============================] - 0s 19ms/step - loss: 2.1498 - accuracy: 0.2933 - val_loss: 1.9184 - val_accuracy: 0.2489
Epoch 4/30
22/22 [==============================] - 0s 21ms/step - loss: 1.5477 - accuracy: 0.3970 - val_loss: 2.1971 - val_accuracy: 0.2800
Epoch 5/30
22/22 [==============================] - 0s 19ms/step - loss: 1.4007 - accuracy: 0.4519 - val_loss: 1.5869 - val_accuracy: 0.3244
Epoch 6/30
22/22 [==============================] - 0s 19ms/step - loss: 0.9661 - accuracy: 0.6148 - val_loss: 1.6954 - val_accuracy: 0.3156
Epoch 7/30
22/22 [==============================] - 0s 19ms/step - loss: 0.6993 - accuracy: 0.7333 - val_loss: 1.6185 - val_accuracy: 0.4222
Epoch 8/30
22/22 [==============================] - 0s 21ms/step - loss: 0.5849 - accuracy: 0.7896 - val_loss: 2.1832 - val_accuracy: 0.3822
Epoch 9/30
22/22 [==============================] - 0s 19ms/step - loss: 0.4039 - accuracy: 0.8756 - val_loss: 1.9544 - val_accuracy: 0.3956
Epoch 10/30
22/22 [==============================] - 0s 19ms/step - loss: 0.2518 - accuracy: 0.9200 - val_loss: 1.9826 - val_accuracy: 0.3911
Epoch 11/30
22/22 [==============================] - 0s 19ms/step - loss: 0.1742 - accuracy: 0.9556 - val_loss: 1.6874 - val_accuracy: 0.4222
Epoch 12/30
22/22 [==============================] - 0s 19ms/step - loss: 0.0867 - accuracy: 0.9867 - val_loss: 2.1390 - val_accuracy: 0.4178
Epoch 13/30
22/22 [==============================] - 0s 18ms/step - loss: 0.0798 - accuracy: 0.9867 - val_loss: 2.3751 - val_accuracy: 0.4044
Epoch 14/30
22/22 [==============================] - 0s 20ms/step - loss: 0.0696 - accuracy: 0.9852 - val_loss: 2.1181 - val_accuracy: 0.3911
Epoch 15/30
22/22 [==============================] - 0s 21ms/step - loss: 0.0216 - accuracy: 0.9985 - val_loss: 3.1490 - val_accuracy: 0.4044
Epoch 16/30
22/22 [==============================] - 0s 19ms/step - loss: 0.0170 - accuracy: 0.9970 - val_loss: 2.5625 - val_accuracy: 0.4356
Epoch 17/30
22/22 [==============================] - 0s 19ms/step - loss: 0.0051 - accuracy: 1.0000 - val_loss: 3.0984 - val_accuracy: 0.4311
Epoch 18/30
22/22 [==============================] - 0s 18ms/step - loss: 0.0026 - accuracy: 1.0000 - val_loss: 3.3215 - val_accuracy: 0.4178
Epoch 19/30
22/22 [==============================] - 0s 19ms/step - loss: 0.0017 - accuracy: 1.0000 - val_loss: 3.1323 - val_accuracy: 0.4133
Epoch 20/30
22/22 [==============================] - 0s 19ms/step - loss: 0.0011 - accuracy: 1.0000 - val_loss: 3.3919 - val_accuracy: 0.4267
Epoch 21/30
22/22 [==============================] - 0s 19ms/step - loss: 8.8288e-04 - accuracy: 1.0000 - val_loss: 3.3471 - val_accuracy: 0.4267
Epoch 22/30
22/22 [==============================] - 0s 19ms/step - loss: 7.6764e-04 - accuracy: 1.0000 - val_loss: 3.4016 - val_accuracy: 0.4267
Epoch 23/30
22/22 [==============================] - 0s 19ms/step - loss: 6.8718e-04 - accuracy: 1.0000 - val_loss: 3.4171 - val_accuracy: 0.4311
Epoch 24/30
22/22 [==============================] - 0s 19ms/step - loss: 6.1976e-04 - accuracy: 1.0000 - val_loss: 3.4653 - val_accuracy: 0.4356
Epoch 25/30
22/22 [==============================] - 0s 19ms/step - loss: 5.6414e-04 - accuracy: 1.0000 - val_loss: 3.5003 - val_accuracy: 0.4444
Epoch 26/30
22/22 [==============================] - 0s 19ms/step - loss: 5.1226e-04 - accuracy: 1.0000 - val_loss: 3.5307 - val_accuracy: 0.4400
Epoch 27/30
22/22 [==============================] - 0s 19ms/step - loss: 4.6924e-04 - accuracy: 1.0000 - val_loss: 3.5348 - val_accuracy: 0.4444
Epoch 28/30
22/22 [==============================] - 0s 19ms/step - loss: 4.3624e-04 - accuracy: 1.0000 - val_loss: 3.5677 - val_accuracy: 0.4400
Epoch 29/30
22/22 [==============================] - 0s 18ms/step - loss: 4.1552e-04 - accuracy: 1.0000 - val_loss: 3.6042 - val_accuracy: 0.4444
Epoch 30/30
22/22 [==============================] - 0s 19ms/step - loss: 3.8174e-04 - accuracy: 1.0000 - val_loss: 3.5995 - val_accuracy: 0.4400

print("################## Training Including Validation ##################")
y_pred_vgg_train = np.argmax(model_Vgg.predict(resized_x_train_total), axis=1)
cm1 = ConfusionMatrix(y_train, y_pred_vgg_train)
PLOT_ConfusionMatrix(cm1,"Training including validation set")
model_vgg_report = classification_report(y_train, y_pred_vgg_train)
print(model_vgg_report)

print("################## Test set ##################")
y_pred_vgg_test = np.argmax(model_Vgg.predict(resized_x_test), axis=1)
cm1 = ConfusionMatrix(y_test, y_pred_vgg_test)
PLOT_ConfusionMatrix(cm1,"Test set")
model_vgg_report = classification_report(y_test, y_pred_vgg_test)
print(model_vgg_report)

################## Training Including Validation ##################

              precision    recall  f1-score   support

           0       0.83      0.77      0.80       180
           1       0.66      0.76      0.71       180
           2       1.00      0.29      0.45       180
           3       0.58      0.86      0.70       180
           4       0.73      0.84      0.78       180

    accuracy                           0.70       900
   macro avg       0.76      0.70      0.69       900
weighted avg       0.76      0.70      0.69       900

################## Test set ##################

              precision    recall  f1-score   support

           0       0.37      0.35      0.36        20
           1       0.29      0.40      0.33        20
           2       0.67      0.10      0.17        20
           3       0.29      0.40      0.33        20
           4       0.32      0.35      0.33        20

    accuracy                           0.32       100
   macro avg       0.38      0.32      0.31       100
weighted avg       0.38      0.32      0.31       100

Part 1b. Transfer Learning - Regression Reformulation (3 Marks)

Step 1 (2)

Repeat the steps of Part 1a. but reformulate as a regression problem, i.e., your network needs to output a single float value ranging between 0 to 6 corresponding to the number of leaves. Again, you are not expected to fine-tune the initial VGG layers.

• Plot a loss curve for training and validation data
• Plot an accuracy curve for training and validation data
• Provide confusion matrix of your network on the training including validation and testing data sets.

Step 2 (1)

The size of the training data is quite small. Discuss based on your learning curves if overfitting is occurring with the models from Parts 1a and 1b.

#Write your code here
start_time = time.time()
import tensorflow
from tensorflow import keras
import warnings
warnings.filterwarnings("ignore")
tensorflow.config.run_functions_eagerly(True)
from sklearn.metrics import accuracy_score
model_Vgg = keras.models.Sequential()

architecture = VGG16(include_top=False, input_shape=(128,128,3), weights='imagenet', classes=5, pooling="avg")

# my coustom metric
def round_preds(y_true, y_pred):
    min = 0
    max = 4
    y_true = y_true.numpy()
    y_pred = y_pred.numpy()

    # print(y_pred)
    y_pred = (y_pred - np.min(y_pred)) / (np.max(y_pred) - np.min(y_pred)) * (max - min) + min
    y_pred = np.round_(y_pred)
    # print(y_pred)
    return accuracy_score(y_true, y_pred)

##################### Freez all layers
for layer in architecture.layers:
    layer.trainable = False
    
for i in range(0,7):
    model_Vgg.add(architecture.layers[i])

# Normalization
resized_x_train = resized_x_train/255
resized_x_val = resized_x_val/255

model_Vgg.add(Conv2D(128, (3, 3), activation="linear",padding='valid')) # layer 1

model_Vgg.add(Conv2D(16, (1, 1), activation=keras.layers.LeakyReLU(),padding='valid')) # 1x1 conv layer
model_Vgg.add(MaxPooling2D(pool_size=(2, 2),padding='valid')) # max pooling layer

model_Vgg.add(Flatten())
model_Vgg.add(Dense(1024, activation=keras.layers.LeakyReLU())) # layer 2
model_Vgg.add(Dense(512, activation=keras.layers.LeakyReLU())) # layer 3
model_Vgg.add(Dense(256, activation=keras.layers.LeakyReLU())) # layer 4
model_Vgg.add(Dense(1, activation="linear")) # layer 5

model_Vgg.compile(loss="mse", optimizer="adam", metrics=[round_preds])

print(model_Vgg.summary())

history_Vgg = model_Vgg.fit(resized_x_train, new_y_train, epochs=400, batch_size = 32, validation_data=(resized_x_val, y_val))

loss_training = history_Vgg.history['loss']
loss_test = history_Vgg.history['val_loss']

accuracy_training = history_Vgg.history["round_preds"]
accuracy_test = history_Vgg.history['val_round_preds']

####################### Plotting
plt.plot(loss_test)
plt.plot(loss_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Loss")
plt.legend(['val Loss', 'Train Loss'])
plt.show()

plt.plot(accuracy_test)
plt.plot(accuracy_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Accuracy")
plt.legend(['val accuracy', 'Train accuracy'])
plt.show()
stop_time = time.time()
print("Total time is :", stop_time - start_time)

Model: "sequential_6"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
block1_conv1 (Conv2D)        (None, 128, 128, 64)      1792      
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 128, 128, 64)      36928     
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 64, 64, 64)        0         
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 64, 64, 128)       73856     
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 64, 64, 128)       147584    
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 32, 32, 128)       0         
_________________________________________________________________
conv2d_12 (Conv2D)           (None, 30, 30, 128)       147584    
_________________________________________________________________
conv2d_13 (Conv2D)           (None, 30, 30, 16)        2064      
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 15, 15, 16)        0         
_________________________________________________________________
flatten_6 (Flatten)          (None, 3600)              0         
_________________________________________________________________
dense_24 (Dense)             (None, 1024)              3687424   
_________________________________________________________________
dense_25 (Dense)             (None, 512)               524800    
_________________________________________________________________
dense_26 (Dense)             (None, 256)               131328    
_________________________________________________________________
dense_27 (Dense)             (None, 1)                 257       
=================================================================
Total params: 4,753,617
Trainable params: 4,493,457
Non-trainable params: 260,160
_________________________________________________________________
None
Epoch 1/200
20/20 [==============================] - 2s 108ms/step - loss: 9444.2373 - round_preds: 0.2209 - val_loss: 2691.4709 - val_round_preds: 0.2525
Epoch 2/200
20/20 [==============================] - 2s 98ms/step - loss: 496.7863 - round_preds: 0.2301 - val_loss: 148.6031 - val_round_preds: 0.2316
Epoch 3/200
20/20 [==============================] - 2s 98ms/step - loss: 38.2284 - round_preds: 0.2199 - val_loss: 20.9774 - val_round_preds: 0.1885
Epoch 4/200
20/20 [==============================] - 2s 97ms/step - loss: 10.6164 - round_preds: 0.2084 - val_loss: 4.3770 - val_round_preds: 0.1652
Epoch 5/200
20/20 [==============================] - 2s 99ms/step - loss: 3.1193 - round_preds: 0.2301 - val_loss: 2.0901 - val_round_preds: 0.2034
Epoch 6/200
20/20 [==============================] - 2s 112ms/step - loss: 2.5616 - round_preds: 0.2301 - val_loss: 2.0254 - val_round_preds: 0.1964
Epoch 7/200
20/20 [==============================] - 2s 96ms/step - loss: 6.3761 - round_preds: 0.2169 - val_loss: 2.0404 - val_round_preds: 0.1905
Epoch 8/200
20/20 [==============================] - 2s 112ms/step - loss: 4.5706 - round_preds: 0.2504 - val_loss: 8.1712 - val_round_preds: 0.2356
Epoch 9/200
20/20 [==============================] - 2s 112ms/step - loss: 4.9540 - round_preds: 0.2347 - val_loss: 5.6823 - val_round_preds: 0.2252
Epoch 10/200
20/20 [==============================] - 2s 96ms/step - loss: 4.6981 - round_preds: 0.2246 - val_loss: 3.1051 - val_round_preds: 0.2426
Epoch 11/200
20/20 [==============================] - 2s 95ms/step - loss: 2.4510 - round_preds: 0.2393 - val_loss: 6.0676 - val_round_preds: 0.2252
Epoch 12/200
20/20 [==============================] - 2s 120ms/step - loss: 3.4276 - round_preds: 0.2301 - val_loss: 2.0823 - val_round_preds: 0.2009
Epoch 13/200
20/20 [==============================] - 2s 97ms/step - loss: 3.6347 - round_preds: 0.2564 - val_loss: 8.5100 - val_round_preds: 0.2401
Epoch 14/200
20/20 [==============================] - 2s 95ms/step - loss: 4.8735 - round_preds: 0.2247 - val_loss: 2.0172 - val_round_preds: 0.2009
Epoch 15/200
20/20 [==============================] - 2s 111ms/step - loss: 4.2600 - round_preds: 0.2317 - val_loss: 2.4380 - val_round_preds: 0.2401
Epoch 16/200
20/20 [==============================] - 2s 111ms/step - loss: 3.0642 - round_preds: 0.2330 - val_loss: 5.0497 - val_round_preds: 0.2192
Epoch 17/200
20/20 [==============================] - 2s 100ms/step - loss: 3.3768 - round_preds: 0.2697 - val_loss: 4.1221 - val_round_preds: 0.2426
Epoch 18/200
20/20 [==============================] - 2s 112ms/step - loss: 3.7305 - round_preds: 0.2551 - val_loss: 4.7962 - val_round_preds: 0.2381
Epoch 19/200
20/20 [==============================] - 2s 96ms/step - loss: 3.1824 - round_preds: 0.2631 - val_loss: 5.4506 - val_round_preds: 0.2287
Epoch 20/200
20/20 [==============================] - 2s 114ms/step - loss: 5.9185 - round_preds: 0.2334 - val_loss: 7.3696 - val_round_preds: 0.2148
Epoch 21/200
20/20 [==============================] - 2s 111ms/step - loss: 2.4057 - round_preds: 0.2659 - val_loss: 1.8855 - val_round_preds: 0.2356
Epoch 22/200
20/20 [==============================] - 2s 114ms/step - loss: 2.7654 - round_preds: 0.2371 - val_loss: 4.1189 - val_round_preds: 0.2540
Epoch 23/200
20/20 [==============================] - 2s 96ms/step - loss: 3.0191 - round_preds: 0.2402 - val_loss: 3.7768 - val_round_preds: 0.2242
Epoch 24/200
20/20 [==============================] - 2s 112ms/step - loss: 2.5962 - round_preds: 0.2776 - val_loss: 3.3845 - val_round_preds: 0.2138
Epoch 25/200
20/20 [==============================] - 2s 112ms/step - loss: 3.0768 - round_preds: 0.2653 - val_loss: 3.0684 - val_round_preds: 0.2371
Epoch 26/200
20/20 [==============================] - 2s 111ms/step - loss: 4.4243 - round_preds: 0.2614 - val_loss: 2.4120 - val_round_preds: 0.2599
Epoch 27/200
20/20 [==============================] - 2s 120ms/step - loss: 2.7667 - round_preds: 0.2886 - val_loss: 4.2164 - val_round_preds: 0.2163
Epoch 28/200
20/20 [==============================] - 2s 96ms/step - loss: 3.2527 - round_preds: 0.2543 - val_loss: 6.3472 - val_round_preds: 0.2540
Epoch 29/200
20/20 [==============================] - 2s 112ms/step - loss: 5.0154 - round_preds: 0.2739 - val_loss: 5.4506 - val_round_preds: 0.2589
Epoch 30/200
20/20 [==============================] - 2s 111ms/step - loss: 5.7417 - round_preds: 0.2774 - val_loss: 3.1792 - val_round_preds: 0.2183
Epoch 31/200
20/20 [==============================] - 2s 98ms/step - loss: 2.4722 - round_preds: 0.2526 - val_loss: 6.2468 - val_round_preds: 0.2252
Epoch 32/200
20/20 [==============================] - 2s 111ms/step - loss: 3.1975 - round_preds: 0.2479 - val_loss: 1.9320 - val_round_preds: 0.1920
Epoch 33/200
20/20 [==============================] - 2s 111ms/step - loss: 2.2151 - round_preds: 0.2963 - val_loss: 2.6398 - val_round_preds: 0.2103
Epoch 34/200
20/20 [==============================] - 2s 97ms/step - loss: 5.0506 - round_preds: 0.2605 - val_loss: 42.9479 - val_round_preds: 0.2366
Epoch 35/200
20/20 [==============================] - 2s 95ms/step - loss: 15.2219 - round_preds: 0.2490 - val_loss: 4.4596 - val_round_preds: 0.2599
Epoch 36/200
20/20 [==============================] - 2s 95ms/step - loss: 3.0577 - round_preds: 0.2692 - val_loss: 1.8859 - val_round_preds: 0.2207
Epoch 37/200
20/20 [==============================] - 2s 111ms/step - loss: 2.3848 - round_preds: 0.2643 - val_loss: 1.8255 - val_round_preds: 0.2183
Epoch 38/200
20/20 [==============================] - 2s 112ms/step - loss: 2.9928 - round_preds: 0.2722 - val_loss: 2.3137 - val_round_preds: 0.2609
Epoch 39/200
20/20 [==============================] - 2s 111ms/step - loss: 2.2267 - round_preds: 0.2989 - val_loss: 1.9558 - val_round_preds: 0.2574
Epoch 40/200
20/20 [==============================] - 2s 111ms/step - loss: 2.8157 - round_preds: 0.2490 - val_loss: 7.2224 - val_round_preds: 0.2564
Epoch 41/200
20/20 [==============================] - 2s 112ms/step - loss: 2.9568 - round_preds: 0.3028 - val_loss: 1.7982 - val_round_preds: 0.2450
Epoch 42/200
20/20 [==============================] - 2s 120ms/step - loss: 1.7150 - round_preds: 0.2871 - val_loss: 2.0772 - val_round_preds: 0.2093
Epoch 43/200
20/20 [==============================] - 2s 112ms/step - loss: 2.7112 - round_preds: 0.2886 - val_loss: 4.6501 - val_round_preds: 0.2475
Epoch 44/200
20/20 [==============================] - 2s 96ms/step - loss: 2.8137 - round_preds: 0.2714 - val_loss: 2.7020 - val_round_preds: 0.2440
Epoch 45/200
20/20 [==============================] - 2s 112ms/step - loss: 2.1243 - round_preds: 0.2760 - val_loss: 1.7950 - val_round_preds: 0.2589
Epoch 46/200
20/20 [==============================] - 2s 112ms/step - loss: 1.8855 - round_preds: 0.3102 - val_loss: 1.7671 - val_round_preds: 0.2609
Epoch 47/200
20/20 [==============================] - 2s 113ms/step - loss: 1.5293 - round_preds: 0.3034 - val_loss: 1.8164 - val_round_preds: 0.2495
Epoch 48/200
20/20 [==============================] - 2s 96ms/step - loss: 2.3137 - round_preds: 0.3051 - val_loss: 1.8534 - val_round_preds: 0.2426
Epoch 49/200
20/20 [==============================] - 2s 111ms/step - loss: 2.3693 - round_preds: 0.3648 - val_loss: 3.0053 - val_round_preds: 0.2530
Epoch 50/200
20/20 [==============================] - 2s 111ms/step - loss: 3.7973 - round_preds: 0.3064 - val_loss: 4.6171 - val_round_preds: 0.2564
Epoch 51/200
20/20 [==============================] - 2s 111ms/step - loss: 3.2908 - round_preds: 0.2925 - val_loss: 2.3967 - val_round_preds: 0.2138
Epoch 52/200
20/20 [==============================] - 2s 111ms/step - loss: 1.8448 - round_preds: 0.3236 - val_loss: 3.0676 - val_round_preds: 0.2252
Epoch 53/200
20/20 [==============================] - 2s 112ms/step - loss: 2.1345 - round_preds: 0.3253 - val_loss: 3.0606 - val_round_preds: 0.2599
Epoch 54/200
20/20 [==============================] - 2s 116ms/step - loss: 2.6101 - round_preds: 0.3223 - val_loss: 2.2852 - val_round_preds: 0.2703
Epoch 55/200
20/20 [==============================] - 2s 96ms/step - loss: 2.5132 - round_preds: 0.3182 - val_loss: 3.6240 - val_round_preds: 0.2659
Epoch 56/200
20/20 [==============================] - 2s 96ms/step - loss: 2.3321 - round_preds: 0.3097 - val_loss: 3.1315 - val_round_preds: 0.2659
Epoch 57/200
20/20 [==============================] - 2s 121ms/step - loss: 2.1298 - round_preds: 0.3426 - val_loss: 1.9650 - val_round_preds: 0.2738
Epoch 58/200
20/20 [==============================] - 2s 99ms/step - loss: 2.7577 - round_preds: 0.3165 - val_loss: 2.4783 - val_round_preds: 0.2227
Epoch 59/200
20/20 [==============================] - 2s 101ms/step - loss: 1.8036 - round_preds: 0.3291 - val_loss: 4.5183 - val_round_preds: 0.2773
Epoch 60/200
20/20 [==============================] - 2s 112ms/step - loss: 1.6429 - round_preds: 0.3362 - val_loss: 2.3756 - val_round_preds: 0.2495
Epoch 61/200
20/20 [==============================] - 2s 112ms/step - loss: 1.7112 - round_preds: 0.3418 - val_loss: 2.3674 - val_round_preds: 0.2738
Epoch 62/200
20/20 [==============================] - 2s 97ms/step - loss: 1.8501 - round_preds: 0.3243 - val_loss: 2.3716 - val_round_preds: 0.2634
Epoch 63/200
20/20 [==============================] - 2s 113ms/step - loss: 1.5245 - round_preds: 0.3486 - val_loss: 2.4210 - val_round_preds: 0.2495
Epoch 64/200
20/20 [==============================] - 2s 104ms/step - loss: 3.1758 - round_preds: 0.3457 - val_loss: 1.7972 - val_round_preds: 0.2148
Epoch 65/200
20/20 [==============================] - 2s 113ms/step - loss: 2.0859 - round_preds: 0.3446 - val_loss: 2.0617 - val_round_preds: 0.2450
Epoch 66/200
20/20 [==============================] - 2s 113ms/step - loss: 1.8624 - round_preds: 0.3457 - val_loss: 2.5491 - val_round_preds: 0.2530
Epoch 67/200
20/20 [==============================] - 2s 112ms/step - loss: 1.5181 - round_preds: 0.3567 - val_loss: 1.7890 - val_round_preds: 0.2564
Epoch 68/200
20/20 [==============================] - 2s 112ms/step - loss: 1.3507 - round_preds: 0.3462 - val_loss: 1.9802 - val_round_preds: 0.2634
Epoch 69/200
20/20 [==============================] - 2s 115ms/step - loss: 2.2189 - round_preds: 0.3297 - val_loss: 1.7878 - val_round_preds: 0.2634
Epoch 70/200
20/20 [==============================] - 2s 112ms/step - loss: 1.3687 - round_preds: 0.3531 - val_loss: 1.7133 - val_round_preds: 0.2644
Epoch 71/200
20/20 [==============================] - 2s 101ms/step - loss: 1.3028 - round_preds: 0.3736 - val_loss: 1.7214 - val_round_preds: 0.2505
Epoch 72/200
20/20 [==============================] - 2s 118ms/step - loss: 1.6847 - round_preds: 0.3369 - val_loss: 2.0332 - val_round_preds: 0.2669
Epoch 73/200
20/20 [==============================] - 2s 112ms/step - loss: 1.4950 - round_preds: 0.3767 - val_loss: 1.7527 - val_round_preds: 0.2589
Epoch 74/200
20/20 [==============================] - 2s 104ms/step - loss: 1.6198 - round_preds: 0.3517 - val_loss: 3.1702 - val_round_preds: 0.2450
Epoch 75/200
20/20 [==============================] - 2s 112ms/step - loss: 1.8687 - round_preds: 0.3693 - val_loss: 1.8095 - val_round_preds: 0.2634
Epoch 76/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2389 - round_preds: 0.3751 - val_loss: 1.8796 - val_round_preds: 0.2495
Epoch 77/200
20/20 [==============================] - 2s 96ms/step - loss: 1.4356 - round_preds: 0.3392 - val_loss: 1.9290 - val_round_preds: 0.2416
Epoch 78/200
20/20 [==============================] - 2s 96ms/step - loss: 2.7334 - round_preds: 0.3439 - val_loss: 1.8980 - val_round_preds: 0.2302
Epoch 79/200
20/20 [==============================] - 2s 95ms/step - loss: 1.5708 - round_preds: 0.3472 - val_loss: 2.6743 - val_round_preds: 0.2014
Epoch 80/200
20/20 [==============================] - 2s 116ms/step - loss: 1.6644 - round_preds: 0.3767 - val_loss: 1.7106 - val_round_preds: 0.2416
Epoch 81/200
20/20 [==============================] - 2s 112ms/step - loss: 1.4625 - round_preds: 0.3722 - val_loss: 2.4143 - val_round_preds: 0.2609
Epoch 82/200
20/20 [==============================] - 2s 112ms/step - loss: 1.4251 - round_preds: 0.3543 - val_loss: 1.8104 - val_round_preds: 0.2574
Epoch 83/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2046 - round_preds: 0.3790 - val_loss: 1.7286 - val_round_preds: 0.2545
Epoch 84/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2453 - round_preds: 0.3595 - val_loss: 3.7608 - val_round_preds: 0.2545
Epoch 85/200
20/20 [==============================] - 2s 117ms/step - loss: 1.8646 - round_preds: 0.3898 - val_loss: 2.0910 - val_round_preds: 0.2371
Epoch 86/200
20/20 [==============================] - 2s 120ms/step - loss: 1.2102 - round_preds: 0.3692 - val_loss: 2.0536 - val_round_preds: 0.2381
Epoch 87/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2851 - round_preds: 0.3784 - val_loss: 1.7604 - val_round_preds: 0.2460
Epoch 88/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9916 - round_preds: 0.3666 - val_loss: 1.7127 - val_round_preds: 0.2381
Epoch 89/200
20/20 [==============================] - 2s 96ms/step - loss: 1.2675 - round_preds: 0.3884 - val_loss: 1.8201 - val_round_preds: 0.2302
Epoch 90/200
20/20 [==============================] - 2s 116ms/step - loss: 1.1507 - round_preds: 0.4124 - val_loss: 1.7418 - val_round_preds: 0.2242
Epoch 91/200
20/20 [==============================] - 2s 115ms/step - loss: 1.2088 - round_preds: 0.4023 - val_loss: 3.5681 - val_round_preds: 0.2599
Epoch 92/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2107 - round_preds: 0.4011 - val_loss: 1.9061 - val_round_preds: 0.2346
Epoch 93/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2787 - round_preds: 0.3979 - val_loss: 1.9449 - val_round_preds: 0.2436
Epoch 94/200
20/20 [==============================] - 2s 96ms/step - loss: 1.0478 - round_preds: 0.4134 - val_loss: 1.9903 - val_round_preds: 0.2163
Epoch 95/200
20/20 [==============================] - 2s 117ms/step - loss: 1.0146 - round_preds: 0.3902 - val_loss: 2.7111 - val_round_preds: 0.2555
Epoch 96/200
20/20 [==============================] - 2s 112ms/step - loss: 1.0618 - round_preds: 0.3659 - val_loss: 2.0384 - val_round_preds: 0.2312
Epoch 97/200
20/20 [==============================] - 2s 97ms/step - loss: 1.0316 - round_preds: 0.4328 - val_loss: 1.8991 - val_round_preds: 0.2495
Epoch 98/200
20/20 [==============================] - 2s 112ms/step - loss: 1.2240 - round_preds: 0.4391 - val_loss: 3.7578 - val_round_preds: 0.2644
Epoch 99/200
20/20 [==============================] - 2s 112ms/step - loss: 2.3142 - round_preds: 0.4033 - val_loss: 2.4517 - val_round_preds: 0.2391
Epoch 100/200
20/20 [==============================] - 2s 117ms/step - loss: 1.2309 - round_preds: 0.3962 - val_loss: 1.7021 - val_round_preds: 0.2173
Epoch 101/200
20/20 [==============================] - 2s 117ms/step - loss: 0.8512 - round_preds: 0.4477 - val_loss: 1.7258 - val_round_preds: 0.2207
Epoch 102/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9182 - round_preds: 0.4476 - val_loss: 1.7307 - val_round_preds: 0.2024
Epoch 103/200
20/20 [==============================] - 2s 112ms/step - loss: 1.0544 - round_preds: 0.4267 - val_loss: 3.5499 - val_round_preds: 0.2024
Epoch 104/200
20/20 [==============================] - 2s 112ms/step - loss: 1.3201 - round_preds: 0.3955 - val_loss: 2.2552 - val_round_preds: 0.2495
Epoch 105/200
20/20 [==============================] - 2s 103ms/step - loss: 0.9533 - round_preds: 0.4484 - val_loss: 1.9367 - val_round_preds: 0.2312
Epoch 106/200
20/20 [==============================] - 2s 112ms/step - loss: 1.0156 - round_preds: 0.4068 - val_loss: 2.2241 - val_round_preds: 0.2336
Epoch 107/200
20/20 [==============================] - 2s 112ms/step - loss: 1.1057 - round_preds: 0.4313 - val_loss: 1.6849 - val_round_preds: 0.2163
Epoch 108/200
20/20 [==============================] - 2s 96ms/step - loss: 0.7485 - round_preds: 0.4298 - val_loss: 1.7324 - val_round_preds: 0.2227
Epoch 109/200
20/20 [==============================] - 2s 112ms/step - loss: 0.8986 - round_preds: 0.4517 - val_loss: 1.8447 - val_round_preds: 0.2312
Epoch 110/200
20/20 [==============================] - 2s 101ms/step - loss: 0.7905 - round_preds: 0.4456 - val_loss: 1.7885 - val_round_preds: 0.2460
Epoch 111/200
20/20 [==============================] - 2s 96ms/step - loss: 0.6532 - round_preds: 0.4719 - val_loss: 2.1292 - val_round_preds: 0.2956
Epoch 112/200
20/20 [==============================] - 2s 112ms/step - loss: 1.1942 - round_preds: 0.4463 - val_loss: 3.5555 - val_round_preds: 0.2589
Epoch 113/200
20/20 [==============================] - 2s 112ms/step - loss: 1.5501 - round_preds: 0.4112 - val_loss: 1.6872 - val_round_preds: 0.2321
Epoch 114/200
20/20 [==============================] - 2s 112ms/step - loss: 1.0594 - round_preds: 0.4391 - val_loss: 2.1137 - val_round_preds: 0.2922
Epoch 115/200
20/20 [==============================] - 3s 128ms/step - loss: 1.6340 - round_preds: 0.4330 - val_loss: 3.1778 - val_round_preds: 0.2728
Epoch 116/200
20/20 [==============================] - 2s 114ms/step - loss: 0.9789 - round_preds: 0.4514 - val_loss: 1.6307 - val_round_preds: 0.2336
Epoch 117/200
20/20 [==============================] - 2s 97ms/step - loss: 0.7562 - round_preds: 0.4491 - val_loss: 2.4691 - val_round_preds: 0.2242
Epoch 118/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6945 - round_preds: 0.4526 - val_loss: 2.7899 - val_round_preds: 0.2703
Epoch 119/200
20/20 [==============================] - 2s 97ms/step - loss: 1.0273 - round_preds: 0.4472 - val_loss: 1.6808 - val_round_preds: 0.2277
Epoch 120/200
20/20 [==============================] - 2s 121ms/step - loss: 0.8528 - round_preds: 0.4345 - val_loss: 1.7090 - val_round_preds: 0.2207
Epoch 121/200
20/20 [==============================] - 2s 101ms/step - loss: 0.6765 - round_preds: 0.5109 - val_loss: 1.6577 - val_round_preds: 0.2312
Epoch 122/200
20/20 [==============================] - 2s 97ms/step - loss: 0.5934 - round_preds: 0.5023 - val_loss: 2.1711 - val_round_preds: 0.2416
Epoch 123/200
20/20 [==============================] - 2s 112ms/step - loss: 0.7380 - round_preds: 0.5017 - val_loss: 2.8454 - val_round_preds: 0.2242
Epoch 124/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9169 - round_preds: 0.4722 - val_loss: 1.8305 - val_round_preds: 0.2505
Epoch 125/200
20/20 [==============================] - 2s 115ms/step - loss: 0.6987 - round_preds: 0.4761 - val_loss: 1.6841 - val_round_preds: 0.2277
Epoch 126/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9091 - round_preds: 0.5240 - val_loss: 2.8573 - val_round_preds: 0.2599
Epoch 127/200
20/20 [==============================] - 2s 112ms/step - loss: 1.8672 - round_preds: 0.4494 - val_loss: 2.7901 - val_round_preds: 0.2530
Epoch 128/200
20/20 [==============================] - 2s 96ms/step - loss: 1.2608 - round_preds: 0.4180 - val_loss: 2.4747 - val_round_preds: 0.2540
Epoch 129/200
20/20 [==============================] - 2s 112ms/step - loss: 1.6086 - round_preds: 0.4516 - val_loss: 3.8243 - val_round_preds: 0.2624
Epoch 130/200
20/20 [==============================] - 2s 123ms/step - loss: 1.9517 - round_preds: 0.4270 - val_loss: 3.7671 - val_round_preds: 0.2564
Epoch 131/200
20/20 [==============================] - 2s 96ms/step - loss: 0.9591 - round_preds: 0.4213 - val_loss: 2.3390 - val_round_preds: 0.2768
Epoch 132/200
20/20 [==============================] - 2s 111ms/step - loss: 1.0234 - round_preds: 0.4233 - val_loss: 1.6397 - val_round_preds: 0.2242
Epoch 133/200
20/20 [==============================] - 2s 112ms/step - loss: 1.0253 - round_preds: 0.4585 - val_loss: 2.3562 - val_round_preds: 0.2346
Epoch 134/200
20/20 [==============================] - 2s 97ms/step - loss: 1.2476 - round_preds: 0.4922 - val_loss: 2.1049 - val_round_preds: 0.2669
Epoch 135/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6100 - round_preds: 0.5010 - val_loss: 1.9801 - val_round_preds: 0.2773
Epoch 136/200
20/20 [==============================] - 2s 113ms/step - loss: 0.7150 - round_preds: 0.4845 - val_loss: 2.3112 - val_round_preds: 0.2450
Epoch 137/200
20/20 [==============================] - 2s 113ms/step - loss: 0.7859 - round_preds: 0.5189 - val_loss: 1.7317 - val_round_preds: 0.2485
Epoch 138/200
20/20 [==============================] - 2s 96ms/step - loss: 0.6479 - round_preds: 0.5155 - val_loss: 1.9313 - val_round_preds: 0.3001
Epoch 139/200
20/20 [==============================] - 2s 112ms/step - loss: 0.5851 - round_preds: 0.5226 - val_loss: 1.6490 - val_round_preds: 0.3219
Epoch 140/200
20/20 [==============================] - 2s 112ms/step - loss: 0.7160 - round_preds: 0.5537 - val_loss: 2.3747 - val_round_preds: 0.2852
Epoch 141/200
20/20 [==============================] - 2s 118ms/step - loss: 0.5999 - round_preds: 0.5259 - val_loss: 1.9391 - val_round_preds: 0.2495
Epoch 142/200
20/20 [==============================] - 2s 96ms/step - loss: 0.6425 - round_preds: 0.5041 - val_loss: 2.5343 - val_round_preds: 0.2505
Epoch 143/200
20/20 [==============================] - 2s 115ms/step - loss: 0.7757 - round_preds: 0.5250 - val_loss: 1.9464 - val_round_preds: 0.2688
Epoch 144/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9045 - round_preds: 0.5568 - val_loss: 2.2283 - val_round_preds: 0.2564
Epoch 145/200
20/20 [==============================] - 2s 100ms/step - loss: 1.2313 - round_preds: 0.5141 - val_loss: 2.1060 - val_round_preds: 0.2564
Epoch 146/200
20/20 [==============================] - 2s 119ms/step - loss: 0.6230 - round_preds: 0.5131 - val_loss: 2.6707 - val_round_preds: 0.2852
Epoch 147/200
20/20 [==============================] - 2s 111ms/step - loss: 0.6415 - round_preds: 0.5210 - val_loss: 1.7328 - val_round_preds: 0.2768
Epoch 148/200
20/20 [==============================] - 2s 113ms/step - loss: 0.4821 - round_preds: 0.5506 - val_loss: 1.7129 - val_round_preds: 0.2679
Epoch 149/200
20/20 [==============================] - 2s 113ms/step - loss: 0.3807 - round_preds: 0.5804 - val_loss: 1.8366 - val_round_preds: 0.2738
Epoch 150/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6145 - round_preds: 0.5990 - val_loss: 1.6633 - val_round_preds: 0.2748
Epoch 151/200
20/20 [==============================] - 2s 115ms/step - loss: 0.4295 - round_preds: 0.5602 - val_loss: 1.7671 - val_round_preds: 0.2495
Epoch 152/200
20/20 [==============================] - 2s 113ms/step - loss: 0.4493 - round_preds: 0.6162 - val_loss: 1.7618 - val_round_preds: 0.2654
Epoch 153/200
20/20 [==============================] - 2s 96ms/step - loss: 0.5610 - round_preds: 0.5700 - val_loss: 2.2056 - val_round_preds: 0.2793
Epoch 154/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6094 - round_preds: 0.5602 - val_loss: 1.6641 - val_round_preds: 0.2669
Epoch 155/200
20/20 [==============================] - 2s 113ms/step - loss: 0.7879 - round_preds: 0.5852 - val_loss: 2.1728 - val_round_preds: 0.2966
Epoch 156/200
20/20 [==============================] - 2s 118ms/step - loss: 0.7449 - round_preds: 0.6139 - val_loss: 1.6344 - val_round_preds: 0.3051
Epoch 157/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4932 - round_preds: 0.6182 - val_loss: 1.7995 - val_round_preds: 0.2510
Epoch 158/200
20/20 [==============================] - 2s 96ms/step - loss: 0.3337 - round_preds: 0.6092 - val_loss: 2.0313 - val_round_preds: 0.2416
Epoch 159/200
20/20 [==============================] - 2s 106ms/step - loss: 0.4825 - round_preds: 0.6080 - val_loss: 1.8788 - val_round_preds: 0.2887
Epoch 160/200
20/20 [==============================] - 2s 97ms/step - loss: 0.4968 - round_preds: 0.6341 - val_loss: 2.7551 - val_round_preds: 0.2783
Epoch 161/200
20/20 [==============================] - 2s 117ms/step - loss: 0.4029 - round_preds: 0.6077 - val_loss: 1.7445 - val_round_preds: 0.2495
Epoch 162/200
20/20 [==============================] - 2s 114ms/step - loss: 0.2714 - round_preds: 0.6634 - val_loss: 1.7690 - val_round_preds: 0.2426
Epoch 163/200
20/20 [==============================] - 2s 112ms/step - loss: 0.2833 - round_preds: 0.6911 - val_loss: 3.0425 - val_round_preds: 0.2912
Epoch 164/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6811 - round_preds: 0.6754 - val_loss: 2.7172 - val_round_preds: 0.2659
Epoch 165/200
20/20 [==============================] - 2s 112ms/step - loss: 0.7761 - round_preds: 0.6230 - val_loss: 1.7555 - val_round_preds: 0.2460
Epoch 166/200
20/20 [==============================] - 2s 115ms/step - loss: 0.4691 - round_preds: 0.5835 - val_loss: 3.4560 - val_round_preds: 0.2877
Epoch 167/200
20/20 [==============================] - 2s 96ms/step - loss: 0.8105 - round_preds: 0.5845 - val_loss: 1.7915 - val_round_preds: 0.2728
Epoch 168/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4387 - round_preds: 0.5783 - val_loss: 1.7698 - val_round_preds: 0.2550
Epoch 169/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4060 - round_preds: 0.6301 - val_loss: 1.9658 - val_round_preds: 0.2738
Epoch 170/200
20/20 [==============================] - 2s 96ms/step - loss: 0.4189 - round_preds: 0.6713 - val_loss: 3.8131 - val_round_preds: 0.2688
Epoch 171/200
20/20 [==============================] - 2s 118ms/step - loss: 0.5578 - round_preds: 0.5686 - val_loss: 1.9782 - val_round_preds: 0.2426
Epoch 172/200
20/20 [==============================] - 2s 112ms/step - loss: 0.6488 - round_preds: 0.5737 - val_loss: 2.1045 - val_round_preds: 0.2495
Epoch 173/200
20/20 [==============================] - 2s 98ms/step - loss: 0.4360 - round_preds: 0.5929 - val_loss: 1.8035 - val_round_preds: 0.2932
Epoch 174/200
20/20 [==============================] - 2s 103ms/step - loss: 1.0015 - round_preds: 0.6101 - val_loss: 1.9503 - val_round_preds: 0.2634
Epoch 175/200
20/20 [==============================] - 2s 96ms/step - loss: 0.4594 - round_preds: 0.6014 - val_loss: 2.0544 - val_round_preds: 0.2634
Epoch 176/200
20/20 [==============================] - 2s 115ms/step - loss: 0.4273 - round_preds: 0.5804 - val_loss: 1.8762 - val_round_preds: 0.2679
Epoch 177/200
20/20 [==============================] - 2s 113ms/step - loss: 0.5083 - round_preds: 0.6224 - val_loss: 1.9711 - val_round_preds: 0.2550
Epoch 178/200
20/20 [==============================] - 2s 96ms/step - loss: 1.1889 - round_preds: 0.6094 - val_loss: 1.7599 - val_round_preds: 0.2564
Epoch 179/200
20/20 [==============================] - 2s 112ms/step - loss: 0.8545 - round_preds: 0.5540 - val_loss: 1.8056 - val_round_preds: 0.2555
Epoch 180/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4488 - round_preds: 0.6075 - val_loss: 1.7837 - val_round_preds: 0.2450
Epoch 181/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4816 - round_preds: 0.6474 - val_loss: 1.8035 - val_round_preds: 0.2321
Epoch 182/200
20/20 [==============================] - 2s 112ms/step - loss: 0.9981 - round_preds: 0.5751 - val_loss: 1.9720 - val_round_preds: 0.2416
Epoch 183/200
20/20 [==============================] - 2s 96ms/step - loss: 0.8526 - round_preds: 0.5531 - val_loss: 2.2229 - val_round_preds: 0.2679
Epoch 184/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4905 - round_preds: 0.6091 - val_loss: 1.6813 - val_round_preds: 0.2426
Epoch 185/200
20/20 [==============================] - 2s 112ms/step - loss: 0.3705 - round_preds: 0.6599 - val_loss: 2.3267 - val_round_preds: 0.2530
Epoch 186/200
20/20 [==============================] - 2s 112ms/step - loss: 0.3412 - round_preds: 0.6854 - val_loss: 1.7356 - val_round_preds: 0.2540
Epoch 187/200
20/20 [==============================] - 2s 113ms/step - loss: 0.4953 - round_preds: 0.7020 - val_loss: 1.7997 - val_round_preds: 0.2564
Epoch 188/200
20/20 [==============================] - 2s 115ms/step - loss: 0.4607 - round_preds: 0.6670 - val_loss: 1.9966 - val_round_preds: 0.2460
Epoch 189/200
20/20 [==============================] - 2s 97ms/step - loss: 0.4723 - round_preds: 0.6479 - val_loss: 2.0630 - val_round_preds: 0.2827
Epoch 190/200
20/20 [==============================] - 2s 112ms/step - loss: 0.4827 - round_preds: 0.6601 - val_loss: 1.9650 - val_round_preds: 0.2356
Epoch 191/200
20/20 [==============================] - 2s 112ms/step - loss: 0.2721 - round_preds: 0.7124 - val_loss: 1.8668 - val_round_preds: 0.2505
Epoch 192/200
20/20 [==============================] - 2s 100ms/step - loss: 0.1831 - round_preds: 0.7317 - val_loss: 1.9156 - val_round_preds: 0.2644
Epoch 193/200
20/20 [==============================] - 2s 96ms/step - loss: 0.1788 - round_preds: 0.7861 - val_loss: 2.2764 - val_round_preds: 0.2679
Epoch 194/200
20/20 [==============================] - 2s 112ms/step - loss: 0.2197 - round_preds: 0.7447 - val_loss: 1.8302 - val_round_preds: 0.2312
Epoch 195/200
20/20 [==============================] - 2s 97ms/step - loss: 0.1784 - round_preds: 0.7923 - val_loss: 2.0774 - val_round_preds: 0.2530
Epoch 196/200
20/20 [==============================] - 2s 113ms/step - loss: 0.2196 - round_preds: 0.8081 - val_loss: 1.7907 - val_round_preds: 0.2475
Epoch 197/200
20/20 [==============================] - 2s 117ms/step - loss: 0.1092 - round_preds: 0.8362 - val_loss: 1.9103 - val_round_preds: 0.2669
Epoch 198/200
20/20 [==============================] - 2s 112ms/step - loss: 0.1359 - round_preds: 0.8729 - val_loss: 1.9506 - val_round_preds: 0.2589
Epoch 199/200
20/20 [==============================] - 2s 112ms/step - loss: 0.1379 - round_preds: 0.8626 - val_loss: 1.8859 - val_round_preds: 0.2748
Epoch 200/200
20/20 [==============================] - 2s 112ms/step - loss: 0.1517 - round_preds: 0.8018 - val_loss: 1.8651 - val_round_preds: 0.2564

Total time is : 429.8443212509155

print("################## Training Including Validation ##################")
y_pred_vgg_train = model_Vgg.predict(resized_x_train_total)
min = 0
max = 4
y_pred_vgg_train = (y_pred_vgg_train - np.min(y_pred_vgg_train)) / (np.max(y_pred_vgg_train) - np.min(y_pred_vgg_train)) * (max - min) + min
y_pred_vgg_train = np.round_(y_pred_vgg_train)
cm1 = ConfusionMatrix(y_train, y_pred_vgg_train)
PLOT_ConfusionMatrix(cm1,"Training including validation set")
model_vgg_report = classification_report(y_train, y_pred_vgg_train)
print(model_vgg_report)

print("################## Test set ##################")
y_pred_vgg_test = model_Vgg.predict(resized_x_test)
min = 0
max = 4
y_pred_vgg_test = (y_pred_vgg_test - np.min(y_pred_vgg_test)) / (np.max(y_pred_vgg_test) - np.min(y_pred_vgg_test)) * (max - min) + min
y_pred_vgg_test = np.round_(y_pred_vgg_test)
cm1 = ConfusionMatrix(y_test, y_pred_vgg_test)
PLOT_ConfusionMatrix(cm1,"Test set")
model_vgg_report = classification_report(y_test, y_pred_vgg_test)
print(model_vgg_report)

################## Training Including Validation ##################

              precision    recall  f1-score   support

           0       0.70      0.09      0.16       180
           1       0.23      0.71      0.34       180
           2       0.13      0.19      0.15       180
           3       0.35      0.09      0.14       180
           4       1.00      0.01      0.01       180

    accuracy                           0.22       900
   macro avg       0.48      0.22      0.16       900
weighted avg       0.48      0.22      0.16       900

################## Test set ##################

              precision    recall  f1-score   support

           0       0.21      0.15      0.18        20
           1       0.20      0.55      0.29        20
           2       0.17      0.20      0.19        20
           3       0.17      0.05      0.08        20
           4       0.00      0.00      0.00        20

    accuracy                           0.19       100
   macro avg       0.15      0.19      0.15       100
weighted avg       0.15      0.19      0.15       100

Write your discussion here.

Part 1a: I Think there it is considered an underfitting rather than overfitting (70% on train including validation, 32% on test).

Part 1b: but on the second model (regression model) things were a little strange, the model was giving the same output for all inputs (the train accuracy and the test accuracy was not moving), i was trying for days playing with the the model paramters and I have seen different results.

1- I have tried different regression loss functions.

2- I have tried different combinations of activation fuctions of different layers (including output layer).

3- I have tried normalization and with out normalization.

4- I have tried different model architechtures.

5- I have tried different optimizers with different parameters.

6- I have built my own custom metric (to round the floating numbers into integer from (0 to 4)), i used an scaling technique to scale the values of my y_pred to (0 to 4) which have ranges of (-1000 to +1000).

but unfortunately nothing woks, the model is underfitted.

I Would be thankful if you told me a good custom metric for the regression model.

but at the end i sticked to the same archtiture of the classification model to be able to compare between thses two models.

I changed the loss function to MSE and the output layer only.

I Would be thankful if you told me the possible reason for the undefitteng of the regression model in your feedback.

Part 2. Improve your Model (4 Marks)

Regularization and data augmentation are common strategies to deal with small datasets.

Step 1 (1.5)

Incorporate Batch Normalization and Dropout into your design the superior network trained in Part 1. You are not expected to fine-tune the initial VGG layers. Again you will provide the following:

• A loss curve for training and validation data
• An accuracy curve for training and validation data
• A confusion matrix of your network on the training including validation and testing data sets.

Step 2 (1.5)

Train the same model from Step 1, now using data augmentation. Again, please provide the same output metrics as in Step 1.

Step 3 (1)

Discuss based on your learning curves and final metrics in Step 2, how large a improvement can be observed from regularization and data augmentation.

here i have chosen the best model which is the classification model to apply step (1) and step (2) on it.

#Write your code here
######################################################### step(1)
import tensorflow
from tensorflow import keras
model_Vgg = keras.models.Sequential()

architecture = VGG16(include_top=False, input_shape=(128,128,3), weights='imagenet', classes=5, pooling="avg")

##################### Freez all layers
for layer in architecture.layers:
    layer.trainable = False
    
for i in range(0,7):
    model_Vgg.add(architecture.layers[i])
    
# Normalization
resized_x_train = resized_x_train/255
resized_x_val = resized_x_val/255

model_Vgg.add(Conv2D(128, (3, 3), activation="linear",padding='valid')) # layer 1

model_Vgg.add(Conv2D(16, (1, 1), activation=keras.layers.LeakyReLU(),padding='valid')) # 1x1 conv layer
model_Vgg.add(MaxPooling2D(pool_size=(2, 2),padding='valid')) # max pooling layer

model_Vgg.add(Flatten())
model_Vgg.add(Dense(1024, activation=keras.layers.LeakyReLU())) # layer 2
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.5))
model_Vgg.add(Dense(512, activation=keras.layers.LeakyReLU())) # layer 3
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.3))
model_Vgg.add(Dense(256, activation=keras.layers.LeakyReLU())) # layer 4
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.2))
model_Vgg.add(Dense(5, activation='softmax')) # layer 5
model_Vgg.compile(loss='sparse_categorical_crossentropy', optimizer="adam", metrics=['accuracy'])

print(model_Vgg.summary())

history_Vgg = model_Vgg.fit(resized_x_train, new_y_train, epochs=100, batch_size = 32, validation_data=(resized_x_val, y_val))

loss_training = history_Vgg.history['loss']
loss_test = history_Vgg.history['val_loss']

accuracy_training = history_Vgg.history['accuracy']
accuracy_test = history_Vgg.history['val_accuracy']

####################### Plotting
plt.plot(loss_test)
plt.plot(loss_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Loss")
plt.legend(['val Loss', 'Train Loss'])
plt.show()

plt.plot(accuracy_test)
plt.plot(accuracy_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Accuracy")
plt.legend(['val accuracy', 'Train accuracy'])
plt.show()

Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
block1_conv1 (Conv2D)        (None, 128, 128, 64)      1792      
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 128, 128, 64)      36928     
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 64, 64, 64)        0         
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 64, 64, 128)       73856     
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 64, 64, 128)       147584    
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 32, 32, 128)       0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 30, 30, 128)       147584    
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 30, 30, 16)        2064      
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 15, 15, 16)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 3600)              0         
_________________________________________________________________
dense_8 (Dense)              (None, 1024)              3687424   
_________________________________________________________________
batch_normalization_6 (Batch (None, 1024)              4096      
_________________________________________________________________
dropout_6 (Dropout)          (None, 1024)              0         
_________________________________________________________________
dense_9 (Dense)              (None, 512)               524800    
_________________________________________________________________
batch_normalization_7 (Batch (None, 512)               2048      
_________________________________________________________________
dropout_7 (Dropout)          (None, 512)               0         
_________________________________________________________________
dense_10 (Dense)             (None, 256)               131328    
_________________________________________________________________
batch_normalization_8 (Batch (None, 256)               1024      
_________________________________________________________________
dropout_8 (Dropout)          (None, 256)               0         
_________________________________________________________________
dense_11 (Dense)             (None, 5)                 1285      
=================================================================
Total params: 4,761,813
Trainable params: 4,498,069
Non-trainable params: 263,744
_________________________________________________________________
None
Epoch 1/100
20/20 [==============================] - 2s 38ms/step - loss: 2.2284 - accuracy: 0.2841 - val_loss: 40.5830 - val_accuracy: 0.2296
Epoch 2/100
20/20 [==============================] - 0s 23ms/step - loss: 1.2643 - accuracy: 0.5556 - val_loss: 28.8846 - val_accuracy: 0.1778
Epoch 3/100
20/20 [==============================] - 0s 21ms/step - loss: 0.8254 - accuracy: 0.6952 - val_loss: 9.3563 - val_accuracy: 0.1926
Epoch 4/100
20/20 [==============================] - 0s 21ms/step - loss: 0.4705 - accuracy: 0.8206 - val_loss: 11.3795 - val_accuracy: 0.1963
Epoch 5/100
20/20 [==============================] - 0s 21ms/step - loss: 0.3276 - accuracy: 0.8683 - val_loss: 6.4103 - val_accuracy: 0.2444
Epoch 6/100
20/20 [==============================] - 0s 22ms/step - loss: 0.2081 - accuracy: 0.9206 - val_loss: 16.0768 - val_accuracy: 0.1926
Epoch 7/100
20/20 [==============================] - 0s 24ms/step - loss: 0.1554 - accuracy: 0.9413 - val_loss: 6.5902 - val_accuracy: 0.3556
Epoch 8/100
20/20 [==============================] - 0s 24ms/step - loss: 0.1218 - accuracy: 0.9603 - val_loss: 7.8759 - val_accuracy: 0.3704
Epoch 9/100
20/20 [==============================] - 0s 22ms/step - loss: 0.1061 - accuracy: 0.9651 - val_loss: 7.0896 - val_accuracy: 0.2926
Epoch 10/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0544 - accuracy: 0.9810 - val_loss: 4.1138 - val_accuracy: 0.3630
Epoch 11/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0516 - accuracy: 0.9810 - val_loss: 4.1290 - val_accuracy: 0.3481
Epoch 12/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0646 - accuracy: 0.9794 - val_loss: 8.0470 - val_accuracy: 0.2000
Epoch 13/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0461 - accuracy: 0.9841 - val_loss: 6.6784 - val_accuracy: 0.2741
Epoch 14/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0654 - accuracy: 0.9778 - val_loss: 7.1543 - val_accuracy: 0.2815
Epoch 15/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0259 - accuracy: 0.9937 - val_loss: 6.0630 - val_accuracy: 0.2852
Epoch 16/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0483 - accuracy: 0.9841 - val_loss: 5.3294 - val_accuracy: 0.3519
Epoch 17/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0659 - accuracy: 0.9857 - val_loss: 4.3810 - val_accuracy: 0.3889
Epoch 18/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0570 - accuracy: 0.9810 - val_loss: 7.4636 - val_accuracy: 0.2444
Epoch 19/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0565 - accuracy: 0.9778 - val_loss: 7.4886 - val_accuracy: 0.2778
Epoch 20/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0602 - accuracy: 0.9794 - val_loss: 6.9568 - val_accuracy: 0.2963
Epoch 21/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0430 - accuracy: 0.9873 - val_loss: 4.9524 - val_accuracy: 0.3481
Epoch 22/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0408 - accuracy: 0.9857 - val_loss: 5.2838 - val_accuracy: 0.3630
Epoch 23/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0486 - accuracy: 0.9841 - val_loss: 5.2437 - val_accuracy: 0.3815
Epoch 24/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0629 - accuracy: 0.9810 - val_loss: 8.0131 - val_accuracy: 0.2815
Epoch 25/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0374 - accuracy: 0.9857 - val_loss: 7.3079 - val_accuracy: 0.2963
Epoch 26/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0712 - accuracy: 0.9730 - val_loss: 5.2050 - val_accuracy: 0.3037
Epoch 27/100
20/20 [==============================] - 0s 24ms/step - loss: 0.0612 - accuracy: 0.9778 - val_loss: 6.6028 - val_accuracy: 0.3741
Epoch 28/100
20/20 [==============================] - 0s 23ms/step - loss: 0.0317 - accuracy: 0.9921 - val_loss: 4.7907 - val_accuracy: 0.3926
Epoch 29/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0358 - accuracy: 0.9905 - val_loss: 7.5480 - val_accuracy: 0.2889
Epoch 30/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0295 - accuracy: 0.9921 - val_loss: 4.9105 - val_accuracy: 0.4000
Epoch 31/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0232 - accuracy: 0.9921 - val_loss: 5.1073 - val_accuracy: 0.3815
Epoch 32/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0389 - accuracy: 0.9857 - val_loss: 5.9401 - val_accuracy: 0.2667
Epoch 33/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0265 - accuracy: 0.9937 - val_loss: 5.3610 - val_accuracy: 0.3370
Epoch 34/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0125 - accuracy: 0.9952 - val_loss: 5.3961 - val_accuracy: 0.3778
Epoch 35/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0168 - accuracy: 0.9937 - val_loss: 7.2661 - val_accuracy: 0.3333
Epoch 36/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0102 - accuracy: 0.9968 - val_loss: 5.5461 - val_accuracy: 0.4074
Epoch 37/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0278 - accuracy: 0.9889 - val_loss: 7.1280 - val_accuracy: 0.2926
Epoch 38/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0275 - accuracy: 0.9921 - val_loss: 7.5145 - val_accuracy: 0.3259
Epoch 39/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0470 - accuracy: 0.9841 - val_loss: 6.8904 - val_accuracy: 0.3296
Epoch 40/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0319 - accuracy: 0.9873 - val_loss: 4.5436 - val_accuracy: 0.3667
Epoch 41/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0599 - accuracy: 0.9825 - val_loss: 4.3783 - val_accuracy: 0.3704
Epoch 42/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0500 - accuracy: 0.9778 - val_loss: 4.3636 - val_accuracy: 0.3741
Epoch 43/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0703 - accuracy: 0.9825 - val_loss: 6.3750 - val_accuracy: 0.3815
Epoch 44/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0258 - accuracy: 0.9937 - val_loss: 5.1024 - val_accuracy: 0.3556
Epoch 45/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0680 - accuracy: 0.9762 - val_loss: 7.7103 - val_accuracy: 0.2444
Epoch 46/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0556 - accuracy: 0.9794 - val_loss: 8.5517 - val_accuracy: 0.2407
Epoch 47/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0153 - accuracy: 0.9968 - val_loss: 7.4644 - val_accuracy: 0.2593
Epoch 48/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0243 - accuracy: 0.9905 - val_loss: 6.7753 - val_accuracy: 0.2667
Epoch 49/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0184 - accuracy: 0.9921 - val_loss: 7.4000 - val_accuracy: 0.2630
Epoch 50/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0322 - accuracy: 0.9873 - val_loss: 10.3973 - val_accuracy: 0.2333
Epoch 51/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0369 - accuracy: 0.9905 - val_loss: 7.4504 - val_accuracy: 0.2926
Epoch 52/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0248 - accuracy: 0.9937 - val_loss: 5.3562 - val_accuracy: 0.3296
Epoch 53/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0250 - accuracy: 0.9937 - val_loss: 5.5693 - val_accuracy: 0.4037
Epoch 54/100
20/20 [==============================] - 0s 24ms/step - loss: 0.0155 - accuracy: 0.9952 - val_loss: 5.5950 - val_accuracy: 0.3889
Epoch 55/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0160 - accuracy: 0.9937 - val_loss: 5.8918 - val_accuracy: 0.4074
Epoch 56/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0167 - accuracy: 0.9921 - val_loss: 7.0301 - val_accuracy: 0.3889
Epoch 57/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0208 - accuracy: 0.9905 - val_loss: 5.8855 - val_accuracy: 0.3630
Epoch 58/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0083 - accuracy: 1.0000 - val_loss: 6.6449 - val_accuracy: 0.3556
Epoch 59/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0094 - accuracy: 0.9968 - val_loss: 6.0849 - val_accuracy: 0.3407
Epoch 60/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0112 - accuracy: 0.9968 - val_loss: 4.9292 - val_accuracy: 0.3852
Epoch 61/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0026 - accuracy: 1.0000 - val_loss: 5.3245 - val_accuracy: 0.3519
Epoch 62/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0074 - accuracy: 0.9984 - val_loss: 5.8257 - val_accuracy: 0.3222
Epoch 63/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0243 - accuracy: 0.9905 - val_loss: 6.5703 - val_accuracy: 0.3296
Epoch 64/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0208 - accuracy: 0.9921 - val_loss: 8.8464 - val_accuracy: 0.2852
Epoch 65/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0235 - accuracy: 0.9905 - val_loss: 11.0455 - val_accuracy: 0.2481
Epoch 66/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0121 - accuracy: 0.9968 - val_loss: 4.4848 - val_accuracy: 0.3667
Epoch 67/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0260 - accuracy: 0.9937 - val_loss: 6.5816 - val_accuracy: 0.3630
Epoch 68/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0083 - accuracy: 0.9968 - val_loss: 5.8756 - val_accuracy: 0.3741
Epoch 69/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0102 - accuracy: 0.9968 - val_loss: 5.4980 - val_accuracy: 0.3630
Epoch 70/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0108 - accuracy: 0.9937 - val_loss: 7.7942 - val_accuracy: 0.3148
Epoch 71/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0111 - accuracy: 0.9968 - val_loss: 5.3697 - val_accuracy: 0.3333
Epoch 72/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0184 - accuracy: 0.9857 - val_loss: 5.0272 - val_accuracy: 0.3630
Epoch 73/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0129 - accuracy: 0.9952 - val_loss: 6.2113 - val_accuracy: 0.3444
Epoch 74/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0062 - accuracy: 1.0000 - val_loss: 5.6018 - val_accuracy: 0.3741
Epoch 75/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0105 - accuracy: 0.9968 - val_loss: 4.9703 - val_accuracy: 0.3667
Epoch 76/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0043 - accuracy: 0.9984 - val_loss: 5.6683 - val_accuracy: 0.3296
Epoch 77/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0048 - accuracy: 0.9984 - val_loss: 4.5335 - val_accuracy: 0.3630
Epoch 78/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0023 - accuracy: 1.0000 - val_loss: 4.5361 - val_accuracy: 0.3667
Epoch 79/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0039 - accuracy: 1.0000 - val_loss: 5.1593 - val_accuracy: 0.3333
Epoch 80/100
20/20 [==============================] - 0s 24ms/step - loss: 0.0021 - accuracy: 1.0000 - val_loss: 5.5041 - val_accuracy: 0.3444
Epoch 81/100
20/20 [==============================] - 0s 22ms/step - loss: 0.0019 - accuracy: 1.0000 - val_loss: 5.6618 - val_accuracy: 0.3333
Epoch 82/100
20/20 [==============================] - 0s 23ms/step - loss: 0.0088 - accuracy: 0.9968 - val_loss: 5.3755 - val_accuracy: 0.3519
Epoch 83/100
20/20 [==============================] - 0s 25ms/step - loss: 0.0011 - accuracy: 1.0000 - val_loss: 5.3341 - val_accuracy: 0.3407
Epoch 84/100
20/20 [==============================] - 0s 24ms/step - loss: 0.0090 - accuracy: 0.9984 - val_loss: 6.7960 - val_accuracy: 0.2704
Epoch 85/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0057 - accuracy: 0.9984 - val_loss: 5.5314 - val_accuracy: 0.3148
Epoch 86/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0034 - accuracy: 1.0000 - val_loss: 6.8586 - val_accuracy: 0.2963
Epoch 87/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0022 - accuracy: 0.9984 - val_loss: 5.8978 - val_accuracy: 0.2963
Epoch 88/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0104 - accuracy: 0.9968 - val_loss: 8.4106 - val_accuracy: 0.2593
Epoch 89/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0052 - accuracy: 0.9984 - val_loss: 4.8101 - val_accuracy: 0.3593
Epoch 90/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0176 - accuracy: 0.9952 - val_loss: 5.3337 - val_accuracy: 0.3815
Epoch 91/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0085 - accuracy: 1.0000 - val_loss: 7.1365 - val_accuracy: 0.3926
Epoch 92/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0041 - accuracy: 1.0000 - val_loss: 4.8866 - val_accuracy: 0.4333
Epoch 93/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0027 - accuracy: 1.0000 - val_loss: 4.1769 - val_accuracy: 0.4185
Epoch 94/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0073 - accuracy: 0.9984 - val_loss: 4.1483 - val_accuracy: 0.4185
Epoch 95/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0115 - accuracy: 0.9968 - val_loss: 5.9212 - val_accuracy: 0.3852
Epoch 96/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0039 - accuracy: 0.9984 - val_loss: 5.6322 - val_accuracy: 0.3926
Epoch 97/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0044 - accuracy: 0.9984 - val_loss: 5.4172 - val_accuracy: 0.3296
Epoch 98/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0222 - accuracy: 0.9937 - val_loss: 7.6943 - val_accuracy: 0.2815
Epoch 99/100
20/20 [==============================] - 0s 21ms/step - loss: 0.0178 - accuracy: 0.9952 - val_loss: 8.0587 - val_accuracy: 0.2593
Epoch 100/100
20/20 [==============================] - 0s 20ms/step - loss: 0.0243 - accuracy: 0.9921 - val_loss: 9.3389 - val_accuracy: 0.2519

########################################## step(1)
print("################## Training Including Validation ##################")
y_pred_vgg_train = np.argmax(model_Vgg.predict(resized_x_train_total), axis=1)
cm1 = ConfusionMatrix(y_train, y_pred_vgg_train)
PLOT_ConfusionMatrix(cm1,"Training including validation set")
model_vgg_report = classification_report(y_train, y_pred_vgg_train)
print(model_vgg_report)

print("################## Test set ##################")
y_pred_vgg_test = np.argmax(model_Vgg.predict(resized_x_test), axis=1)
cm1 = ConfusionMatrix(y_test, y_pred_vgg_test)
PLOT_ConfusionMatrix(cm1,"Test set")
model_vgg_report = classification_report(y_test, y_pred_vgg_test)
print(model_vgg_report)

################## Training Including Validation ##################

              precision    recall  f1-score   support

           0       0.90      0.48      0.63       180
           1       0.88      0.62      0.73       180
           2       1.00      0.04      0.09       180
           3       0.33      0.97      0.49       180
           4       0.81      0.62      0.70       180

    accuracy                           0.55       900
   macro avg       0.78      0.55      0.53       900
weighted avg       0.78      0.55      0.53       900

################## Test set ##################

              precision    recall  f1-score   support

           0       0.33      0.05      0.09        20
           1       0.33      0.05      0.09        20
           2       0.00      0.00      0.00        20
           3       0.20      0.85      0.32        20
           4       0.33      0.15      0.21        20

    accuracy                           0.22       100
   macro avg       0.24      0.22      0.14       100
weighted avg       0.24      0.22      0.14       100

/opt/conda/lib/python3.7/site-packages/sklearn/metrics/_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
/opt/conda/lib/python3.7/site-packages/sklearn/metrics/_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
/opt/conda/lib/python3.7/site-packages/sklearn/metrics/_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

#Write your code here
######################################################### step(2)
######################################################### Data Augmentation
Img_generator = ImageDataGenerator(rotation_range=90,shear_range=0.5, zoom_range=0.2, horizontal_flip=True)

print(resized_x_train.shape)
print(new_y_train.shape)

for i in range(5):
    iterator = Img_generator.flow(resized_x_train, new_y_train, batch_size=256)

    new_imgs_train = next(iterator)[0].astype("uint8")
    new_labels_train = next(iterator)[1].astype("uint8")

    print(new_imgs_train.shape)
    print(new_labels_train.shape)

    print("******************* Augmented images *******************")
    for i in range(2):
        plt.imshow(new_imgs_train[i])
        plt.show()

    resized_x_train = np.append(resized_x_train, new_imgs_train, axis = 0)
    new_y_train = np.append(new_y_train, new_labels_train, axis = 0)

    print(resized_x_train.shape)
    print(new_y_train.shape)

(630, 128, 128, 3)
(630,)
(256, 128, 128, 3)
(256,)
******************* Augmented images *******************

(886, 128, 128, 3)
(886,)
(256, 128, 128, 3)
(256,)
******************* Augmented images *******************

(1142, 128, 128, 3)
(1142,)
(256, 128, 128, 3)
(256,)
******************* Augmented images *******************

(1398, 128, 128, 3)
(1398,)
(256, 128, 128, 3)
(256,)
******************* Augmented images *******************

(1654, 128, 128, 3)
(1654,)
(256, 128, 128, 3)
(256,)
******************* Augmented images *******************

(1910, 128, 128, 3)
(1910,)

resized_x_train.shape

(1910, 128, 128, 3)

new_y_train.shape

(1910,)

#Write your code here
######################################################### step(2)
import tensorflow
from tensorflow import keras
model_Vgg = keras.models.Sequential()

architecture = VGG16(include_top=False, input_shape=(128,128,3), weights='imagenet', classes=5, pooling="avg")

##################### Freez all layers
for layer in architecture.layers:
    layer.trainable = False
    
for i in range(0,7):
    model_Vgg.add(architecture.layers[i])
    
# Normalization
resized_x_train = resized_x_train/255
resized_x_val = resized_x_val/255

model_Vgg.add(Conv2D(128, (3, 3), activation="linear",padding='valid')) # layer 1

model_Vgg.add(Conv2D(16, (1, 1), activation=keras.layers.LeakyReLU(),padding='valid')) # 1x1 conv layer
model_Vgg.add(MaxPooling2D(pool_size=(2, 2),padding='valid')) # max pooling layer

model_Vgg.add(Flatten())
model_Vgg.add(Dense(1024, activation=keras.layers.LeakyReLU())) # layer 2
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.5))
model_Vgg.add(Dense(512, activation=keras.layers.LeakyReLU())) # layer 3
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.3))
model_Vgg.add(Dense(256, activation=keras.layers.LeakyReLU())) # layer 4
model_Vgg.add(BatchNormalization())
model_Vgg.add(Dropout(0.2))
model_Vgg.add(Dense(5, activation='softmax')) # layer 5
model_Vgg.compile(loss='sparse_categorical_crossentropy', optimizer="adam", metrics=['accuracy'])

print(model_Vgg.summary())

history_Vgg = model_Vgg.fit(resized_x_train, new_y_train, epochs=200, batch_size = 32, validation_data=(resized_x_val, y_val))

loss_training = history_Vgg.history['loss']
loss_test = history_Vgg.history['val_loss']

accuracy_training = history_Vgg.history['accuracy']
accuracy_test = history_Vgg.history['val_accuracy']

####################### Plotting
plt.plot(loss_test)
plt.plot(loss_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Loss")
plt.legend(['val Loss', 'Train Loss'])
plt.show()

plt.plot(accuracy_test)
plt.plot(accuracy_training)
plt.xlabel("# Of Epochs")
plt.ylabel("Accuracy")
plt.legend(['val accuracy', 'Train accuracy'])
plt.show()

Model: "sequential_3"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
block1_conv1 (Conv2D)        (None, 128, 128, 64)      1792      
_________________________________________________________________
block1_conv2 (Conv2D)        (None, 128, 128, 64)      36928     
_________________________________________________________________
block1_pool (MaxPooling2D)   (None, 64, 64, 64)        0         
_________________________________________________________________
block2_conv1 (Conv2D)        (None, 64, 64, 128)       73856     
_________________________________________________________________
block2_conv2 (Conv2D)        (None, 64, 64, 128)       147584    
_________________________________________________________________
block2_pool (MaxPooling2D)   (None, 32, 32, 128)       0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 30, 30, 128)       147584    
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 30, 30, 16)        2064      
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 15, 15, 16)        0         
_________________________________________________________________
flatten_3 (Flatten)          (None, 3600)              0         
_________________________________________________________________
dense_12 (Dense)             (None, 1024)              3687424   
_________________________________________________________________
batch_normalization_9 (Batch (None, 1024)              4096      
_________________________________________________________________
dropout_9 (Dropout)          (None, 1024)              0         
_________________________________________________________________
dense_13 (Dense)             (None, 512)               524800    
_________________________________________________________________
batch_normalization_10 (Batc (None, 512)               2048      
_________________________________________________________________
dropout_10 (Dropout)         (None, 512)               0         
_________________________________________________________________
dense_14 (Dense)             (None, 256)               131328    
_________________________________________________________________
batch_normalization_11 (Batc (None, 256)               1024      
_________________________________________________________________
dropout_11 (Dropout)         (None, 256)               0         
_________________________________________________________________
dense_15 (Dense)             (None, 5)                 1285      
=================================================================
Total params: 4,761,813
Trainable params: 4,498,069
Non-trainable params: 263,744
_________________________________________________________________
None
Epoch 1/200
60/60 [==============================] - 2s 21ms/step - loss: 2.2101 - accuracy: 0.1979 - val_loss: 6.2772 - val_accuracy: 0.1815
Epoch 2/200
60/60 [==============================] - 1s 18ms/step - loss: 1.7612 - accuracy: 0.2712 - val_loss: 1.9203 - val_accuracy: 0.2296
Epoch 3/200
60/60 [==============================] - 1s 17ms/step - loss: 1.6168 - accuracy: 0.3518 - val_loss: 2.3519 - val_accuracy: 0.2667
Epoch 4/200
60/60 [==============================] - 1s 17ms/step - loss: 1.4244 - accuracy: 0.4319 - val_loss: 2.6540 - val_accuracy: 0.2370
Epoch 5/200
60/60 [==============================] - 1s 17ms/step - loss: 1.1977 - accuracy: 0.5309 - val_loss: 2.5905 - val_accuracy: 0.2222
Epoch 6/200
60/60 [==============================] - 1s 17ms/step - loss: 1.0144 - accuracy: 0.6241 - val_loss: 3.2199 - val_accuracy: 0.2259
Epoch 7/200
60/60 [==============================] - 1s 17ms/step - loss: 0.7326 - accuracy: 0.7283 - val_loss: 4.1403 - val_accuracy: 0.2148
Epoch 8/200
60/60 [==============================] - 1s 17ms/step - loss: 0.6143 - accuracy: 0.7911 - val_loss: 4.4815 - val_accuracy: 0.2593
Epoch 9/200
60/60 [==============================] - 1s 17ms/step - loss: 0.4060 - accuracy: 0.8455 - val_loss: 4.4723 - val_accuracy: 0.2556
Epoch 10/200
60/60 [==============================] - 1s 17ms/step - loss: 0.3128 - accuracy: 0.8932 - val_loss: 5.9030 - val_accuracy: 0.2407
Epoch 11/200
60/60 [==============================] - 1s 17ms/step - loss: 0.2914 - accuracy: 0.8984 - val_loss: 6.1781 - val_accuracy: 0.1852
Epoch 12/200
60/60 [==============================] - 1s 17ms/step - loss: 0.2362 - accuracy: 0.9204 - val_loss: 4.4077 - val_accuracy: 0.2778
Epoch 13/200
60/60 [==============================] - 1s 18ms/step - loss: 0.1708 - accuracy: 0.9419 - val_loss: 4.6004 - val_accuracy: 0.2407
Epoch 14/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1568 - accuracy: 0.9508 - val_loss: 6.6262 - val_accuracy: 0.1963
Epoch 15/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1442 - accuracy: 0.9492 - val_loss: 4.7430 - val_accuracy: 0.2370
Epoch 16/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1379 - accuracy: 0.9529 - val_loss: 5.6622 - val_accuracy: 0.2259
Epoch 17/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1104 - accuracy: 0.9613 - val_loss: 7.5347 - val_accuracy: 0.2148
Epoch 18/200
60/60 [==============================] - 1s 19ms/step - loss: 0.1411 - accuracy: 0.9518 - val_loss: 5.6369 - val_accuracy: 0.2704
Epoch 19/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1358 - accuracy: 0.9618 - val_loss: 5.3606 - val_accuracy: 0.2407
Epoch 20/200
60/60 [==============================] - 1s 17ms/step - loss: 0.1057 - accuracy: 0.9654 - val_loss: 5.2423 - val_accuracy: 0.2444
Epoch 21/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0702 - accuracy: 0.9743 - val_loss: 5.3013 - val_accuracy: 0.2593
Epoch 22/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0670 - accuracy: 0.9770 - val_loss: 5.3185 - val_accuracy: 0.2481
Epoch 23/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0786 - accuracy: 0.9707 - val_loss: 5.4761 - val_accuracy: 0.2593
Epoch 24/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0878 - accuracy: 0.9686 - val_loss: 4.6677 - val_accuracy: 0.2889
Epoch 25/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0774 - accuracy: 0.9728 - val_loss: 7.7500 - val_accuracy: 0.2148
Epoch 26/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0620 - accuracy: 0.9832 - val_loss: 5.7489 - val_accuracy: 0.2778
Epoch 27/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0559 - accuracy: 0.9817 - val_loss: 5.2382 - val_accuracy: 0.2667
Epoch 28/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0596 - accuracy: 0.9770 - val_loss: 6.9459 - val_accuracy: 0.2296
Epoch 29/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0753 - accuracy: 0.9728 - val_loss: 5.7158 - val_accuracy: 0.2556
Epoch 30/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0770 - accuracy: 0.9696 - val_loss: 5.8387 - val_accuracy: 0.2630
Epoch 31/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0447 - accuracy: 0.9869 - val_loss: 6.1372 - val_accuracy: 0.2704
Epoch 32/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0495 - accuracy: 0.9822 - val_loss: 6.3612 - val_accuracy: 0.2556
Epoch 33/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0496 - accuracy: 0.9822 - val_loss: 6.4487 - val_accuracy: 0.2333
Epoch 34/200
60/60 [==============================] - 1s 19ms/step - loss: 0.0690 - accuracy: 0.9759 - val_loss: 7.1743 - val_accuracy: 0.2519
Epoch 35/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0706 - accuracy: 0.9770 - val_loss: 7.6905 - val_accuracy: 0.2444
Epoch 36/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0710 - accuracy: 0.9728 - val_loss: 5.8946 - val_accuracy: 0.2407
Epoch 37/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0734 - accuracy: 0.9712 - val_loss: 7.1670 - val_accuracy: 0.2519
Epoch 38/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0685 - accuracy: 0.9775 - val_loss: 6.2173 - val_accuracy: 0.2259
Epoch 39/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0929 - accuracy: 0.9696 - val_loss: 7.2109 - val_accuracy: 0.2111
Epoch 40/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0688 - accuracy: 0.9743 - val_loss: 5.6220 - val_accuracy: 0.2519
Epoch 41/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0757 - accuracy: 0.9743 - val_loss: 6.3384 - val_accuracy: 0.2185
Epoch 42/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0908 - accuracy: 0.9707 - val_loss: 8.0101 - val_accuracy: 0.2481
Epoch 43/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0795 - accuracy: 0.9749 - val_loss: 6.3070 - val_accuracy: 0.2185
Epoch 44/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0983 - accuracy: 0.9675 - val_loss: 6.3252 - val_accuracy: 0.2556
Epoch 45/200
60/60 [==============================] - 1s 19ms/step - loss: 0.0708 - accuracy: 0.9780 - val_loss: 7.7134 - val_accuracy: 0.2074
Epoch 46/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0661 - accuracy: 0.9759 - val_loss: 6.6387 - val_accuracy: 0.2407
Epoch 47/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0276 - accuracy: 0.9906 - val_loss: 6.8767 - val_accuracy: 0.2037
Epoch 48/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0283 - accuracy: 0.9916 - val_loss: 6.4249 - val_accuracy: 0.2148
Epoch 49/200
60/60 [==============================] - 1s 20ms/step - loss: 0.0303 - accuracy: 0.9880 - val_loss: 6.7696 - val_accuracy: 0.2630
Epoch 50/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0435 - accuracy: 0.9832 - val_loss: 7.1280 - val_accuracy: 0.2407
Epoch 51/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0436 - accuracy: 0.9853 - val_loss: 6.9707 - val_accuracy: 0.2185
Epoch 52/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0441 - accuracy: 0.9838 - val_loss: 6.4699 - val_accuracy: 0.2667
Epoch 53/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0372 - accuracy: 0.9890 - val_loss: 8.3817 - val_accuracy: 0.2296
Epoch 54/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0597 - accuracy: 0.9806 - val_loss: 7.4528 - val_accuracy: 0.2148
Epoch 55/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0389 - accuracy: 0.9832 - val_loss: 6.6204 - val_accuracy: 0.2556
Epoch 56/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0421 - accuracy: 0.9853 - val_loss: 6.4634 - val_accuracy: 0.2333
Epoch 57/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0354 - accuracy: 0.9874 - val_loss: 6.4179 - val_accuracy: 0.2259
Epoch 58/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0509 - accuracy: 0.9838 - val_loss: 10.6340 - val_accuracy: 0.2111
Epoch 59/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0389 - accuracy: 0.9911 - val_loss: 7.5421 - val_accuracy: 0.2259
Epoch 60/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0327 - accuracy: 0.9874 - val_loss: 6.7639 - val_accuracy: 0.2296
Epoch 61/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0378 - accuracy: 0.9885 - val_loss: 7.8741 - val_accuracy: 0.2519
Epoch 62/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0322 - accuracy: 0.9911 - val_loss: 7.8198 - val_accuracy: 0.2519
Epoch 63/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0478 - accuracy: 0.9838 - val_loss: 7.1535 - val_accuracy: 0.2407
Epoch 64/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0492 - accuracy: 0.9853 - val_loss: 7.2093 - val_accuracy: 0.2259
Epoch 65/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0473 - accuracy: 0.9838 - val_loss: 8.2487 - val_accuracy: 0.2259
Epoch 66/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0501 - accuracy: 0.9801 - val_loss: 7.4859 - val_accuracy: 0.2111
Epoch 67/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0474 - accuracy: 0.9848 - val_loss: 9.4238 - val_accuracy: 0.2407
Epoch 68/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0393 - accuracy: 0.9859 - val_loss: 8.6494 - val_accuracy: 0.2000
Epoch 69/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0377 - accuracy: 0.9890 - val_loss: 8.7062 - val_accuracy: 0.2259
Epoch 70/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0299 - accuracy: 0.9869 - val_loss: 7.3490 - val_accuracy: 0.2667
Epoch 71/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0465 - accuracy: 0.9843 - val_loss: 7.5515 - val_accuracy: 0.2370
Epoch 72/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0715 - accuracy: 0.9754 - val_loss: 6.8154 - val_accuracy: 0.2815
Epoch 73/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0561 - accuracy: 0.9843 - val_loss: 8.6321 - val_accuracy: 0.2556
Epoch 74/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0659 - accuracy: 0.9759 - val_loss: 8.5789 - val_accuracy: 0.2185
Epoch 75/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0657 - accuracy: 0.9754 - val_loss: 7.1185 - val_accuracy: 0.2296
Epoch 76/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0376 - accuracy: 0.9901 - val_loss: 6.7611 - val_accuracy: 0.2481
Epoch 77/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0450 - accuracy: 0.9817 - val_loss: 8.7154 - val_accuracy: 0.2407
Epoch 78/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0347 - accuracy: 0.9895 - val_loss: 7.6835 - val_accuracy: 0.2407
Epoch 79/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0329 - accuracy: 0.9880 - val_loss: 6.2038 - val_accuracy: 0.2630
Epoch 80/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0380 - accuracy: 0.9874 - val_loss: 6.9452 - val_accuracy: 0.2222
Epoch 81/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0316 - accuracy: 0.9880 - val_loss: 7.2302 - val_accuracy: 0.2630
Epoch 82/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0431 - accuracy: 0.9853 - val_loss: 8.4063 - val_accuracy: 0.2556
Epoch 83/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0433 - accuracy: 0.9848 - val_loss: 8.1463 - val_accuracy: 0.2296
Epoch 84/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0586 - accuracy: 0.9853 - val_loss: 7.6837 - val_accuracy: 0.2593
Epoch 85/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0487 - accuracy: 0.9838 - val_loss: 8.2108 - val_accuracy: 0.2259
Epoch 86/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0482 - accuracy: 0.9822 - val_loss: 7.6138 - val_accuracy: 0.2444
Epoch 87/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0544 - accuracy: 0.9806 - val_loss: 8.4138 - val_accuracy: 0.2407
Epoch 88/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0530 - accuracy: 0.9822 - val_loss: 8.6948 - val_accuracy: 0.2481
Epoch 89/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0386 - accuracy: 0.9869 - val_loss: 7.8630 - val_accuracy: 0.2296
Epoch 90/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0573 - accuracy: 0.9848 - val_loss: 7.2753 - val_accuracy: 0.2630
Epoch 91/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0410 - accuracy: 0.9864 - val_loss: 7.7228 - val_accuracy: 0.2556
Epoch 92/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0226 - accuracy: 0.9921 - val_loss: 8.1615 - val_accuracy: 0.2222
Epoch 93/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0255 - accuracy: 0.9921 - val_loss: 7.4532 - val_accuracy: 0.2444
Epoch 94/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0333 - accuracy: 0.9890 - val_loss: 9.7386 - val_accuracy: 0.2074
Epoch 95/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0264 - accuracy: 0.9911 - val_loss: 8.0423 - val_accuracy: 0.2556
Epoch 96/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0274 - accuracy: 0.9906 - val_loss: 6.7008 - val_accuracy: 0.2778
Epoch 97/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0298 - accuracy: 0.9911 - val_loss: 6.8000 - val_accuracy: 0.3000
Epoch 98/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0267 - accuracy: 0.9901 - val_loss: 7.5623 - val_accuracy: 0.2296
Epoch 99/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0296 - accuracy: 0.9911 - val_loss: 8.2007 - val_accuracy: 0.2259
Epoch 100/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0319 - accuracy: 0.9911 - val_loss: 8.1251 - val_accuracy: 0.2704
Epoch 101/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0313 - accuracy: 0.9895 - val_loss: 8.1425 - val_accuracy: 0.2481
Epoch 102/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0330 - accuracy: 0.9885 - val_loss: 6.3432 - val_accuracy: 0.2778
Epoch 103/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0340 - accuracy: 0.9890 - val_loss: 7.5230 - val_accuracy: 0.2519
Epoch 104/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0349 - accuracy: 0.9895 - val_loss: 8.0436 - val_accuracy: 0.2444
Epoch 105/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0325 - accuracy: 0.9885 - val_loss: 9.0001 - val_accuracy: 0.2111
Epoch 106/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0356 - accuracy: 0.9869 - val_loss: 8.3128 - val_accuracy: 0.2407
Epoch 107/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0351 - accuracy: 0.9890 - val_loss: 8.8736 - val_accuracy: 0.2704
Epoch 108/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0339 - accuracy: 0.9874 - val_loss: 9.4140 - val_accuracy: 0.2222
Epoch 109/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0205 - accuracy: 0.9916 - val_loss: 8.3651 - val_accuracy: 0.2556
Epoch 110/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0446 - accuracy: 0.9874 - val_loss: 8.3491 - val_accuracy: 0.2407
Epoch 111/200
60/60 [==============================] - 1s 19ms/step - loss: 0.0246 - accuracy: 0.9895 - val_loss: 6.9337 - val_accuracy: 0.2778
Epoch 112/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0259 - accuracy: 0.9937 - val_loss: 8.4453 - val_accuracy: 0.2370
Epoch 113/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0242 - accuracy: 0.9927 - val_loss: 8.1538 - val_accuracy: 0.2296
Epoch 114/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0226 - accuracy: 0.9906 - val_loss: 8.7772 - val_accuracy: 0.2259
Epoch 115/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0277 - accuracy: 0.9911 - val_loss: 8.4753 - val_accuracy: 0.2333
Epoch 116/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0352 - accuracy: 0.9864 - val_loss: 9.0175 - val_accuracy: 0.2000
Epoch 117/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0384 - accuracy: 0.9885 - val_loss: 7.3821 - val_accuracy: 0.2556
Epoch 118/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0280 - accuracy: 0.9885 - val_loss: 7.3569 - val_accuracy: 0.2704
Epoch 119/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0424 - accuracy: 0.9853 - val_loss: 9.5187 - val_accuracy: 0.2370
Epoch 120/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0378 - accuracy: 0.9869 - val_loss: 7.3329 - val_accuracy: 0.2630
Epoch 121/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0319 - accuracy: 0.9880 - val_loss: 7.5575 - val_accuracy: 0.2444
Epoch 122/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0329 - accuracy: 0.9911 - val_loss: 8.7675 - val_accuracy: 0.2407
Epoch 123/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0285 - accuracy: 0.9911 - val_loss: 8.4096 - val_accuracy: 0.2556
Epoch 124/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0462 - accuracy: 0.9864 - val_loss: 7.8482 - val_accuracy: 0.2519
Epoch 125/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0293 - accuracy: 0.9911 - val_loss: 9.5197 - val_accuracy: 0.1963
Epoch 126/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0172 - accuracy: 0.9953 - val_loss: 10.6148 - val_accuracy: 0.2074
Epoch 127/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0073 - accuracy: 0.9990 - val_loss: 8.2532 - val_accuracy: 0.2111
Epoch 128/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0117 - accuracy: 0.9963 - val_loss: 7.6698 - val_accuracy: 0.2630
Epoch 129/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0132 - accuracy: 0.9969 - val_loss: 8.5472 - val_accuracy: 0.2630
Epoch 130/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0143 - accuracy: 0.9953 - val_loss: 8.2025 - val_accuracy: 0.2556
Epoch 131/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0215 - accuracy: 0.9927 - val_loss: 8.1024 - val_accuracy: 0.2630
Epoch 132/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0382 - accuracy: 0.9859 - val_loss: 8.2348 - val_accuracy: 0.1963
Epoch 133/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0281 - accuracy: 0.9906 - val_loss: 8.6391 - val_accuracy: 0.2259
Epoch 134/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0351 - accuracy: 0.9885 - val_loss: 11.7351 - val_accuracy: 0.2037
Epoch 135/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0360 - accuracy: 0.9880 - val_loss: 7.5591 - val_accuracy: 0.2370
Epoch 136/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0260 - accuracy: 0.9921 - val_loss: 8.9332 - val_accuracy: 0.2074
Epoch 137/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0302 - accuracy: 0.9901 - val_loss: 8.7114 - val_accuracy: 0.2630
Epoch 138/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0385 - accuracy: 0.9880 - val_loss: 8.3096 - val_accuracy: 0.2926
Epoch 139/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0555 - accuracy: 0.9827 - val_loss: 7.9319 - val_accuracy: 0.2444
Epoch 140/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0441 - accuracy: 0.9848 - val_loss: 7.1572 - val_accuracy: 0.2815
Epoch 141/200
60/60 [==============================] - 1s 19ms/step - loss: 0.0499 - accuracy: 0.9832 - val_loss: 9.3357 - val_accuracy: 0.2333
Epoch 142/200
60/60 [==============================] - 1s 20ms/step - loss: 0.0487 - accuracy: 0.9843 - val_loss: 7.6989 - val_accuracy: 0.2370
Epoch 143/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0411 - accuracy: 0.9859 - val_loss: 8.0506 - val_accuracy: 0.2444
Epoch 144/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0288 - accuracy: 0.9885 - val_loss: 8.6897 - val_accuracy: 0.2481
Epoch 145/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0261 - accuracy: 0.9895 - val_loss: 9.8264 - val_accuracy: 0.2185
Epoch 146/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0107 - accuracy: 0.9969 - val_loss: 7.8386 - val_accuracy: 0.2296
Epoch 147/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0060 - accuracy: 0.9990 - val_loss: 8.0632 - val_accuracy: 0.2370
Epoch 148/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0062 - accuracy: 0.9979 - val_loss: 8.2952 - val_accuracy: 0.2333
Epoch 149/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0062 - accuracy: 0.9995 - val_loss: 8.0713 - val_accuracy: 0.2296
Epoch 150/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0060 - accuracy: 0.9984 - val_loss: 7.3402 - val_accuracy: 0.2444
Epoch 151/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0042 - accuracy: 0.9990 - val_loss: 7.9853 - val_accuracy: 0.2407
Epoch 152/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0103 - accuracy: 0.9963 - val_loss: 8.8304 - val_accuracy: 0.2407
Epoch 153/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0149 - accuracy: 0.9969 - val_loss: 7.7746 - val_accuracy: 0.2593
Epoch 154/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0154 - accuracy: 0.9942 - val_loss: 9.5333 - val_accuracy: 0.2407
Epoch 155/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0153 - accuracy: 0.9948 - val_loss: 8.9581 - val_accuracy: 0.2296
Epoch 156/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0126 - accuracy: 0.9963 - val_loss: 9.7606 - val_accuracy: 0.2259
Epoch 157/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0153 - accuracy: 0.9969 - val_loss: 8.9245 - val_accuracy: 0.2593
Epoch 158/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0251 - accuracy: 0.9906 - val_loss: 8.7648 - val_accuracy: 0.1963
Epoch 159/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0405 - accuracy: 0.9853 - val_loss: 8.4675 - val_accuracy: 0.2630
Epoch 160/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0220 - accuracy: 0.9932 - val_loss: 11.1585 - val_accuracy: 0.2259
Epoch 161/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0063 - accuracy: 0.9990 - val_loss: 7.9327 - val_accuracy: 0.2519
Epoch 162/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0227 - accuracy: 0.9911 - val_loss: 9.4510 - val_accuracy: 0.2667
Epoch 163/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0235 - accuracy: 0.9927 - val_loss: 9.0816 - val_accuracy: 0.2185
Epoch 164/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0243 - accuracy: 0.9916 - val_loss: 9.8764 - val_accuracy: 0.2444
Epoch 165/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0428 - accuracy: 0.9859 - val_loss: 7.9837 - val_accuracy: 0.2333
Epoch 166/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0226 - accuracy: 0.9901 - val_loss: 8.6388 - val_accuracy: 0.2222
Epoch 167/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0281 - accuracy: 0.9895 - val_loss: 10.7714 - val_accuracy: 0.2185
Epoch 168/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0358 - accuracy: 0.9885 - val_loss: 9.2113 - val_accuracy: 0.2333
Epoch 169/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0239 - accuracy: 0.9921 - val_loss: 9.6287 - val_accuracy: 0.2296
Epoch 170/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0202 - accuracy: 0.9937 - val_loss: 9.8159 - val_accuracy: 0.1926
Epoch 171/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0175 - accuracy: 0.9958 - val_loss: 8.8896 - val_accuracy: 0.2333
Epoch 172/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0177 - accuracy: 0.9921 - val_loss: 9.5405 - val_accuracy: 0.2444
Epoch 173/200
60/60 [==============================] - 1s 22ms/step - loss: 0.0122 - accuracy: 0.9958 - val_loss: 8.2135 - val_accuracy: 0.2778
Epoch 174/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0177 - accuracy: 0.9932 - val_loss: 9.7219 - val_accuracy: 0.2407
Epoch 175/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0115 - accuracy: 0.9932 - val_loss: 9.2081 - val_accuracy: 0.2630
Epoch 176/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0148 - accuracy: 0.9942 - val_loss: 9.4991 - val_accuracy: 0.2556
Epoch 177/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0105 - accuracy: 0.9974 - val_loss: 8.8239 - val_accuracy: 0.2370
Epoch 178/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0108 - accuracy: 0.9953 - val_loss: 8.0449 - val_accuracy: 0.2667
Epoch 179/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0150 - accuracy: 0.9953 - val_loss: 8.5395 - val_accuracy: 0.2222
Epoch 180/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0119 - accuracy: 0.9948 - val_loss: 8.4675 - val_accuracy: 0.2630
Epoch 181/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0123 - accuracy: 0.9953 - val_loss: 9.0671 - val_accuracy: 0.2444
Epoch 182/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0220 - accuracy: 0.9937 - val_loss: 9.6083 - val_accuracy: 0.2481
Epoch 183/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0244 - accuracy: 0.9921 - val_loss: 9.3091 - val_accuracy: 0.2481
Epoch 184/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0468 - accuracy: 0.9853 - val_loss: 9.2000 - val_accuracy: 0.2630
Epoch 185/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0272 - accuracy: 0.9885 - val_loss: 9.6748 - val_accuracy: 0.2481
Epoch 186/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0114 - accuracy: 0.9963 - val_loss: 9.3982 - val_accuracy: 0.2333
Epoch 187/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0259 - accuracy: 0.9906 - val_loss: 7.7351 - val_accuracy: 0.2370
Epoch 188/200
60/60 [==============================] - 1s 16ms/step - loss: 0.0423 - accuracy: 0.9869 - val_loss: 8.9870 - val_accuracy: 0.1963
Epoch 189/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0311 - accuracy: 0.9864 - val_loss: 9.9383 - val_accuracy: 0.2333
Epoch 190/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0198 - accuracy: 0.9927 - val_loss: 9.5604 - val_accuracy: 0.2111
Epoch 191/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0367 - accuracy: 0.9880 - val_loss: 8.2163 - val_accuracy: 0.2556
Epoch 192/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0344 - accuracy: 0.9911 - val_loss: 8.4629 - val_accuracy: 0.2407
Epoch 193/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0212 - accuracy: 0.9916 - val_loss: 8.7020 - val_accuracy: 0.2556
Epoch 194/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0277 - accuracy: 0.9901 - val_loss: 8.9758 - val_accuracy: 0.2519
Epoch 195/200
60/60 [==============================] - 1s 18ms/step - loss: 0.0192 - accuracy: 0.9932 - val_loss: 9.1250 - val_accuracy: 0.2556
Epoch 196/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0115 - accuracy: 0.9963 - val_loss: 9.1759 - val_accuracy: 0.2481
Epoch 197/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0115 - accuracy: 0.9969 - val_loss: 8.6791 - val_accuracy: 0.2667
Epoch 198/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0111 - accuracy: 0.9958 - val_loss: 10.2252 - val_accuracy: 0.2370
Epoch 199/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0104 - accuracy: 0.9963 - val_loss: 8.5708 - val_accuracy: 0.2111
Epoch 200/200
60/60 [==============================] - 1s 17ms/step - loss: 0.0169 - accuracy: 0.9942 - val_loss: 8.5415 - val_accuracy: 0.2481

########################################## step(1)
print("################## Training Including Validation ##################")
y_pred_vgg_train = np.argmax(model_Vgg.predict(resized_x_train_total), axis=1)
cm1 = ConfusionMatrix(y_train, y_pred_vgg_train)
PLOT_ConfusionMatrix(cm1,"Training including validation set")
model_vgg_report = classification_report(y_train, y_pred_vgg_train)
print(model_vgg_report)

print("################## Test set ##################")
y_pred_vgg_test = np.argmax(model_Vgg.predict(resized_x_test), axis=1)
cm1 = ConfusionMatrix(y_test, y_pred_vgg_test)
PLOT_ConfusionMatrix(cm1,"Test set")
model_vgg_report = classification_report(y_test, y_pred_vgg_test)
print(model_vgg_report)

################## Training Including Validation ##################

              precision    recall  f1-score   support

           0       0.73      0.59      0.65       180
           1       0.72      0.65      0.68       180
           2       0.90      0.44      0.59       180
           3       0.47      0.78      0.58       180
           4       0.60      0.67      0.63       180

    accuracy                           0.63       900
   macro avg       0.68      0.63      0.63       900
weighted avg       0.68      0.63      0.63       900

################## Test set ##################

              precision    recall  f1-score   support

           0       0.30      0.15      0.20        20
           1       0.17      0.25      0.20        20
           2       0.40      0.20      0.27        20
           3       0.24      0.45      0.32        20
           4       0.21      0.15      0.18        20

    accuracy                           0.24       100
   macro avg       0.27      0.24      0.23       100
weighted avg       0.27      0.24      0.23       100

Write your discussion here.

for these 2 steps i have tried the same architecture of the best model which was the classification model.

but unfortunately the test accuracy decreased a little bit, but in step (1) the gap between the test accuracy and the train accuracy decreased a little bit.

and in step (2) the training accuracy is higher than the training accuracy in step (1).

but still the first model is the best model which was the classification model that has 70% on tain and validation and 32% on test.

but unfortunately the (best model) after reguralization and data augmentation perform worse than the first best classification model.

References

[1] N. Teimouri, M. Dyrmann, P. R. Nielsen, S. K. Mathiassen, G. J. Somerville, and R. N. Jørgensen, “Weed growth stage estimator using deep convolutional neural networks,” Sensors, vol. 18, no. 5, 2018.